abs
stringlengths 44
64
| Download PDF
stringlengths 75
115
| OpenReview
stringlengths 42
42
| title
stringlengths 15
148
| url
stringlengths 44
64
| authors
stringlengths 6
903
| detail_url
stringlengths 44
64
| tags
stringclasses 1
value | abstract
stringlengths 422
5.84k
|
|---|---|---|---|---|---|---|---|---|
https://proceedings.mlr.press/v235/chen24bc.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chen24bc/chen24bc.pdf | https://openreview.net/forum?id=vSerUPYFtB | One for All: A Universal Generator for Concept Unlearnability via Multi-Modal Alignment | https://proceedings.mlr.press/v235/chen24bc.html | Chaochao Chen, Jiaming Zhang, Yuyuan Li, Zhongxuan Han | https://proceedings.mlr.press/v235/chen24bc.html | ICML 2024 | The abundance of free internet data offers unprecedented opportunities for researchers and developers, but it also poses privacy risks. Utilizing data without explicit consent raises critical challenges in protecting personal information.Unlearnable examples have emerged as a feasible protection approach, which renders the data unlearnable, i.e., useless to third parties, by injecting imperceptible perturbations. However, these perturbations only exhibit unlearnable effects on either a particular dataset or label-consistent scenarios, thereby lacking broad applicability. To address both issues concurrently, we propose a universal perturbation generator that harnesses data with concept unlearnability, thereby broadening the scope of unlearnability beyond specific datasets or labels. Specifically, we leverage multi-modal pre-trained models to establish a connection between the data concepts in a shared embedding space. This connection enables the information transformation from image data to text concepts. Consequently, we can align the text embedding using concept-wise discriminant loss, and render the data unlearnable. Extensive experiments conducted on real-world datasets demonstrate the concept unlearnability, i.e., cross-dataset transferability and label-agnostic utility, of our proposed unlearnable examples, as well as their robustness against attacks. |
https://proceedings.mlr.press/v235/chen24bd.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chen24bd/chen24bd.pdf | https://openreview.net/forum?id=ohG9bVMs5j | Generating In-Distribution Proxy Graphs for Explaining Graph Neural Networks | https://proceedings.mlr.press/v235/chen24bd.html | Zhuomin Chen, Jiaxing Zhang, Jingchao Ni, Xiaoting Li, Yuchen Bian, Md Mezbahul Islam, Ananda Mondal, Hua Wei, Dongsheng Luo | https://proceedings.mlr.press/v235/chen24bd.html | ICML 2024 | Graph Neural Networks (GNNs) have become a building block in graph data processing, with wide applications in critical domains. The growing needs to deploy GNNs in high-stakes applications necessitate explainability for users in the decision-making processes. A popular paradigm for the explainability of GNNs is to identify explainable subgraphs by comparing their labels with the ones of original graphs. This task is challenging due to the substantial distributional shift from the original graphs in the training set to the set of explainable subgraphs, which prevents accurate prediction of labels with the subgraphs. To address it, in this paper, we propose a novel method that generates proxy graphs for explainable subgraphs that are in the distribution of training data. We introduce a parametric method that employs graph generators to produce proxy graphs. A new training objective based on information theory is designed to ensure that proxy graphs not only adhere to the distribution of training data but also preserve explanatory factors. Such generated proxy graphs can be reliably used to approximate the predictions of the labels of explainable subgraphs. Empirical evaluations across various datasets demonstrate our method achieves more accurate explanations for GNNs. |
https://proceedings.mlr.press/v235/chen24be.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chen24be/chen24be.pdf | https://openreview.net/forum?id=QH4mXDEULp | Diffusive Gibbs Sampling | https://proceedings.mlr.press/v235/chen24be.html | Wenlin Chen, Mingtian Zhang, Brooks Paige, José Miguel Hernández-Lobato, David Barber | https://proceedings.mlr.press/v235/chen24be.html | ICML 2024 | The inadequate mixing of conventional Markov Chain Monte Carlo (MCMC) methods for multi-modal distributions presents a significant challenge in practical applications such as Bayesian inference and molecular dynamics. Addressing this, we propose Diffusive Gibbs Sampling (DiGS), an innovative family of sampling methods designed for effective sampling from distributions characterized by distant and disconnected modes. DiGS integrates recent developments in diffusion models, leveraging Gaussian convolution to create an auxiliary noisy distribution that bridges isolated modes in the original space and applying Gibbs sampling to alternately draw samples from both spaces. A novel Metropolis-within-Gibbs scheme is proposed to enhance mixing in the denoising sampling step. DiGS exhibits a better mixing property for sampling multi-modal distributions than state-of-the-art methods such as parallel tempering, attaining substantially improved performance across various tasks, including mixtures of Gaussians, Bayesian neural networks and molecular dynamics. |
https://proceedings.mlr.press/v235/chen24bf.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chen24bf/chen24bf.pdf | https://openreview.net/forum?id=0xmfExPqFf | Provable Risk-Sensitive Distributional Reinforcement Learning with General Function Approximation | https://proceedings.mlr.press/v235/chen24bf.html | Yu Chen, Xiangcheng Zhang, Siwei Wang, Longbo Huang | https://proceedings.mlr.press/v235/chen24bf.html | ICML 2024 | In the realm of reinforcement learning (RL), accounting for risk is crucial for making decisions under uncertainty, particularly in applications where safety and reliability are paramount. In this paper, we introduce a general framework on Risk-Sensitive Distributional Reinforcement Learning (RS-DisRL), with static Lipschitz Risk Measures (LRM) and general function approximation. Our framework covers a broad class of risk-sensitive RL, and facilitates analysis of the impact of estimation functions on the effectiveness of RSRL strategies and evaluation of their sample complexity. We design two innovative meta-algorithms: RS-DisRL-M, a model-based strategy for model-based function approximation, and RS-DisRL-V, a model-free approach for general value function approximation. With our novel estimation techniques via Least Squares Regression (LSR) and Maximum Likelihood Estimation (MLE) in distributional RL with augmented Markov Decision Process (MDP), we derive the first $\widetilde{\mathcal{O}}(\sqrt{K})$ dependency of the regret upper bound for RSRL with static LRM, marking a pioneering contribution towards statistically efficient algorithms in this domain. |
https://proceedings.mlr.press/v235/chen24bg.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chen24bg/chen24bg.pdf | https://openreview.net/forum?id=LyJ85kgHFe | $\textttMoE-RBench$: Towards Building Reliable Language Models with Sparse Mixture-of-Experts | https://proceedings.mlr.press/v235/chen24bg.html | Guanjie Chen, Xinyu Zhao, Tianlong Chen, Yu Cheng | https://proceedings.mlr.press/v235/chen24bg.html | ICML 2024 | Mixture-of-Experts (MoE) has gained increasing popularity as a promising framework for scaling up large language models (LLMs). However, the reliability assessment of MoE lags behind its surging applications. Moreover, when transferred to new domains such as in fine-tuning MoE models sometimes underperform their dense counterparts. Motivated by the research gap and counter-intuitive phenomenon, we propose $\texttt{MoE-RBench}$, the first comprehensive assessment of SMoE reliability from three aspects: $\textit{(i)}$ safety and hallucination, $\textit{(ii)}$ resilience to adversarial attacks, and $\textit{(iii)}$ out-of-distribution robustness. Extensive models and datasets are tested to compare the MoE to dense networks from these reliability dimensions. Our empirical observations suggest that with appropriate hyperparameters, training recipes, and inference techniques, we can build the MoE model more reliably than the dense LLM. In particular, we find that the robustness of SMoE is sensitive to the basic training settings. We hope that this study can provide deeper insights into how to adapt the pre-trained MoE model to other tasks with higher-generation security, quality, and stability. Codes are available at https://github.com/UNITES-Lab/MoE-RBench. |
https://proceedings.mlr.press/v235/chen24bh.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chen24bh/chen24bh.pdf | https://openreview.net/forum?id=B48Pzc4oKi | LLaGA: Large Language and Graph Assistant | https://proceedings.mlr.press/v235/chen24bh.html | Runjin Chen, Tong Zhao, Ajay Kumar Jaiswal, Neil Shah, Zhangyang Wang | https://proceedings.mlr.press/v235/chen24bh.html | ICML 2024 | Graph Neural Networks (GNNs) have empowered the advance in graph-structured data analysis. Recently, the rise of Large Language Models (LLMs) like GPT-4 has heralded a new era in deep learning. However, their application to graph data poses distinct challenges due to the inherent difficulty of translating graph structures to language. To this end, we introduce the the Large Language and Graph Assistant (LLaGA), an innovative model that effectively integrates LLM capabilities to handle the complexities of graph-structured data. LLaGA retains the general-purpose nature of LLMs while adapting graph data into a format compatible with LLM input. LLaGA achieves this by reorganizing graph nodes to structure-aware sequences and then mapping these into the token embedding space through a versatile projector. LLaGA excels in versatility, generalizability and interpretability, allowing it to perform consistently well across different datasets and tasks, extend its ability to unseen datasets or tasks, and provide explanations for graphs. Our extensive experiments across popular graph benchmarks show that LLaGA delivers outstanding performance across four datasets and three tasks using one single model, surpassing state-of-the-art graph models in both supervised and zero-shot scenarios. |
https://proceedings.mlr.press/v235/chen24bi.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chen24bi/chen24bi.pdf | https://openreview.net/forum?id=EYvEVbfoDp | HALC: Object Hallucination Reduction via Adaptive Focal-Contrast Decoding | https://proceedings.mlr.press/v235/chen24bi.html | Zhaorun Chen, Zhuokai Zhao, Hongyin Luo, Huaxiu Yao, Bo Li, Jiawei Zhou | https://proceedings.mlr.press/v235/chen24bi.html | ICML 2024 | While large vision-language models (LVLMs) have demonstrated impressive capabilities in interpreting multi-modal contexts, they invariably suffer from object hallucinations (OH). We introduce HALC, a novel decoding algorithm designed to mitigate OH in LVLMs. HALC leverages distinct fine-grained optimal visual information in vision-language tasks and operates on both local and global contexts simultaneously. Specifically, HALC integrates a robust auto-focal grounding mechanism (locally) to correct hallucinated tokens on the fly, and a specialized beam search algorithm (globally) to significantly reduce OH while preserving text generation quality. Additionally, HALC can be integrated into any LVLMs as a plug-and-play module without extra training. Extensive experimental studies demonstrate HALC’s effectiveness in reducing OH, outperforming state-of-the-arts across four benchmarks. Code is released at https://github.com/BillChan226/HALC. |
https://proceedings.mlr.press/v235/chen24bj.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chen24bj/chen24bj.pdf | https://openreview.net/forum?id=OdsZS0E0AO | Compact Optimality Verification for Optimization Proxies | https://proceedings.mlr.press/v235/chen24bj.html | Wenbo Chen, Haoruo Zhao, Mathieu Tanneau, Pascal Van Hentenryck | https://proceedings.mlr.press/v235/chen24bj.html | ICML 2024 | Recent years have witnessed increasing interest in optimization proxies, i.e., machine learning models that approximate the input-output mapping of parametric optimization problems and return near-optimal feasible solutions. Following recent work by (Nellikkath & Chatzivasileiadis, 2021), this paper reconsiders the optimality verification problem for optimization proxies, i.e., the determination of the worst-case optimality gap over the instance distribution. The paper proposes a compact formulation for optimality verification and a gradient-based primal heuristic that brings significant computational benefits to the original formulation. The compact formulation is also more general and applies to non-convex optimization problems. The benefits of the compact formulation are demonstrated on large-scale DC Optimal Power Flow and knapsack problems. |
https://proceedings.mlr.press/v235/chen24bk.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chen24bk/chen24bk.pdf | https://openreview.net/forum?id=XBNhJQU84y | Enhancing Implicit Shape Generators Using Topological Regularizations | https://proceedings.mlr.press/v235/chen24bk.html | Liyan Chen, Yan Zheng, Yang Li, Lohit Anirudh Jagarapu, Haoxiang Li, Hao Kang, Gang Hua, Qixing Huang | https://proceedings.mlr.press/v235/chen24bk.html | ICML 2024 | A fundamental problem in learning 3D shapes generative models is that when the generative model is simply fitted to the training data, the resulting synthetic 3D models can present various artifacts. Many of these artifacts are topological in nature, e.g., broken legs, unrealistic thin structures, and small holes. In this paper, we introduce a principled approach that utilizes topological regularization losses on an implicit shape generator to rectify topological artifacts. The objectives are two-fold. The first is to align the persistent diagram (PD) distribution of the training shapes with that of synthetic shapes. The second ensures that the PDs are smooth among adjacent synthetic shapes. We show how to achieve these two objectives using two simple but effective formulations. Specifically, distribution alignment is achieved to learn a generative model of PDs and align this generator with PDs of synthetic shapes. We show how to handle discrete and continuous variabilities of PDs by using a shape-regularization term when performing PD alignment. Moreover, we enforce the smoothness of the PDs using a smoothness loss on the PD generator, which further improves the behavior of PD distribution alignment. Experimental results on ShapeNet show that our approach leads to much better generalization behavior than state-of-the-art implicit shape generators. |
https://proceedings.mlr.press/v235/chen24bl.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chen24bl/chen24bl.pdf | https://openreview.net/forum?id=99jx5U81jx | Do Models Explain Themselves? Counterfactual Simulatability of Natural Language Explanations | https://proceedings.mlr.press/v235/chen24bl.html | Yanda Chen, Ruiqi Zhong, Narutatsu Ri, Chen Zhao, He He, Jacob Steinhardt, Zhou Yu, Kathleen Mckeown | https://proceedings.mlr.press/v235/chen24bl.html | ICML 2024 | Large language models (LLMs) are trained to imitate humans to explain human decisions. However, do LLMs explain themselves? Can they help humans build mental models of how LLMs process different inputs? To answer these questions, we propose to evaluate $\textbf{counterfactual simulatability}$ of natural language explanations: whether an explanation can enable humans to precisely infer the model’s outputs on diverse counterfactuals of the explained input. For example, if a model answers ”$\textit{yes}$” to the input question ”$\textit{Can eagles fly?}$” with the explanation ”$\textit{all birds can fly}$”, then humans would infer from the explanation that it would also answer ”$\textit{yes}$” to the counterfactual input ”$\textit{Can penguins fly?}$”. If the explanation is precise, then the model’s answer should match humans’ expectations. We implemented two metrics based on counterfactual simulatability: precision and generality. We generated diverse counterfactuals automatically using LLMs. We then used these metrics to evaluate state-of-the-art LLMs (e.g., GPT-4) on two tasks: multi-hop factual reasoning and reward modeling. We found that LLM’s explanations have low precision and that precision does not correlate with plausibility. Therefore, naively optimizing human approvals (e.g., RLHF) may be insufficient. |
https://proceedings.mlr.press/v235/chen24bm.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chen24bm/chen24bm.pdf | https://openreview.net/forum?id=H86WzfH5N1 | On the Trajectory Regularity of ODE-based Diffusion Sampling | https://proceedings.mlr.press/v235/chen24bm.html | Defang Chen, Zhenyu Zhou, Can Wang, Chunhua Shen, Siwei Lyu | https://proceedings.mlr.press/v235/chen24bm.html | ICML 2024 | Diffusion-based generative models use stochastic differential equations (SDEs) and their equivalent ordinary differential equations (ODEs) to establish a smooth connection between a complex data distribution and a tractable prior distribution. In this paper, we identify several intriguing trajectory properties in the ODE-based sampling process of diffusion models. We characterize an implicit denoising trajectory and discuss its vital role in forming the coupled sampling trajectory with a strong shape regularity, regardless of the generated content. We also describe a dynamic programming-based scheme to make the time schedule in sampling better fit the underlying trajectory structure. This simple strategy requires minimal modification to any given ODE-based numerical solvers and incurs negligible computational cost, while delivering superior performance in image generation, especially in $5\sim 10$ function evaluations. |
https://proceedings.mlr.press/v235/chen24bn.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chen24bn/chen24bn.pdf | https://openreview.net/forum?id=zcIV8OQFVF | ODIN: Disentangled Reward Mitigates Hacking in RLHF | https://proceedings.mlr.press/v235/chen24bn.html | Lichang Chen, Chen Zhu, Jiuhai Chen, Davit Soselia, Tianyi Zhou, Tom Goldstein, Heng Huang, Mohammad Shoeybi, Bryan Catanzaro | https://proceedings.mlr.press/v235/chen24bn.html | ICML 2024 | In this work, we study the issue of reward hacking on the response length, a challenge emerging in Reinforcement Learning from Human Feedback (RLHF) on LLMs. A well-formatted, verbose but less helpful response from the LLMs can often deceive LLMs or even human evaluators and achieve high scores. The same issue also holds for some reward models in RL. To address the challenges in both training and evaluation, we establish a more reliable evaluation protocol for comparing different training configurations, which inspects the trade-off between LLM evaluation score and response length obtained by varying training hyperparameters. Based on this evaluation, we conduct large-scale studies, where the results shed insights into the efficacy of hyperparameters and tricks used in RL on mitigating length bias. We further propose to improve the reward model by jointly training two linear heads to predict the preference, one trained to correlate with length and the other trained to decorrelate with length and therefore focusing more on the actual content. We then discard the length head in RL to ignore the spurious length reward. Experiments demonstrate that our approach eliminates the reward correlation with length, and improves the obtained policy by a significant margin. |
https://proceedings.mlr.press/v235/chen24bo.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chen24bo/chen24bo.pdf | https://openreview.net/forum?id=Lgq1E92h1U | Stacking Deep Set Networks and Pooling by Quantiles | https://proceedings.mlr.press/v235/chen24bo.html | Zhuojun Chen, Xinghua Zhu, Dongzhe Su, Justin C. I. Chuang | https://proceedings.mlr.press/v235/chen24bo.html | ICML 2024 | We propose Stacked Deep Sets and Quantile Pooling for learning tasks on set data. We introduce Quantile Pooling, a novel permutation-invariant pooling operation that synergizes max and average pooling. Just like max pooling, quantile pooling emphasizes the most salient features of the data. Like average pooling, it captures the overall distribution and subtle features of the data. Like both, it is lightweight and fast. We demonstrate the effectiveness of our approach in a variety of tasks, showing that quantile pooling can outperform both max and average pooling in each of their respective strengths. We also introduce a variant of deep set networks that is more expressive and universal. While Quantile Pooling balances robustness and sensitivity, Stacked Deep Sets enhances learning with depth. |
https://proceedings.mlr.press/v235/chen24bp.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chen24bp/chen24bp.pdf | https://openreview.net/forum?id=YNbCbcGyXE | What Can Transformer Learn with Varying Depth? Case Studies on Sequence Learning Tasks | https://proceedings.mlr.press/v235/chen24bp.html | Xingwu Chen, Difan Zou | https://proceedings.mlr.press/v235/chen24bp.html | ICML 2024 | We study the capabilities of the transformer architecture with varying depth. Specifically, we designed a novel set of sequence learning tasks to systematically evaluate and comprehend how the depth of transformer affects its ability to perform memorization, reasoning, generalization, and contextual generalization. We show a transformer with only one attention layer can excel in memorization but falls short in other tasks. Then, we show that exhibiting reasoning and generalization ability requires the transformer to have at least two attention layers, while context generalization ability may necessitate three attention layers. Additionally, we identify a class of simple operations that a single attention layer can execute, and show that the complex tasks can be approached as the combinations of these simple operations and thus can be resolved by stacking multiple attention layers. This sheds light on studying more practical and complex tasks beyond our design. Numerical experiments corroborate our theoretical findings. |
https://proceedings.mlr.press/v235/cheng24a.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/cheng24a/cheng24a.pdf | https://openreview.net/forum?id=ah1BlQcLv4 | Transformers Implement Functional Gradient Descent to Learn Non-Linear Functions In Context | https://proceedings.mlr.press/v235/cheng24a.html | Xiang Cheng, Yuxin Chen, Suvrit Sra | https://proceedings.mlr.press/v235/cheng24a.html | ICML 2024 | Many neural network architectures are known to be Turing Complete, and can thus, in principle implement arbitrary algorithms. However, Transformers are unique in that they can implement gradient-based learning algorithms under simple parameter configurations. This paper provides theoretical and empirical evidence that (non-linear) Transformers naturally learn to implement gradient descent in function space, which in turn enable them to learn non-linear functions in context. Our results apply to a broad class of combinations of non-linear architectures and non-linear in-context learning tasks. Additionally, we show that the optimal choice of non-linear activation depends in a natural way on the class of functions that need to be learned. |
https://proceedings.mlr.press/v235/cheng24b.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/cheng24b/cheng24b.pdf | https://openreview.net/forum?id=7zEoinErzQ | Layerwise Change of Knowledge in Neural Networks | https://proceedings.mlr.press/v235/cheng24b.html | Xu Cheng, Lei Cheng, Zhaoran Peng, Yang Xu, Tian Han, Quanshi Zhang | https://proceedings.mlr.press/v235/cheng24b.html | ICML 2024 | This paper aims to explain how a deep neural network (DNN) gradually extracts new knowledge and forgets noisy features through layers in forward propagation. Up to now, although how to define knowledge encoded by the DNN has not reached a consensus so far, previous studies have derived a series of mathematical evidences to take interactions as symbolic primitive inference patterns encoded by a DNN. We extend the definition of interactions and, for the first time, extract interactions encoded by intermediate layers. We quantify and track the newly emerged interactions and the forgotten interactions in each layer during the forward propagation, which shed new light on the learning behavior of DNNs. The layer-wise change of interactions also reveals the change of the generalization capacity and instability of feature representations of a DNN. |
https://proceedings.mlr.press/v235/cheng24c.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/cheng24c/cheng24c.pdf | https://openreview.net/forum?id=CecY6XiUfu | Reference Neural Operators: Learning the Smooth Dependence of Solutions of PDEs on Geometric Deformations | https://proceedings.mlr.press/v235/cheng24c.html | Ze Cheng, Zhongkai Hao, Xiaoqiang Wang, Jianing Huang, Youjia Wu, Xudan Liu, Yiru Zhao, Songming Liu, Hang Su | https://proceedings.mlr.press/v235/cheng24c.html | ICML 2024 | For partial differential equations on domains of arbitrary shapes, existing works of neural operators attempt to learn a mapping from geometries to solutions. It often requires a large dataset of geometry-solution pairs in order to obtain a sufficiently accurate neural operator. However, for many industrial applications, e.g., engineering design optimization, it can be prohibitive to satisfy the requirement since even a single simulation may take hours or days of computation. To address this issue, we propose reference neural operators (RNO), a novel way of implementing neural operators, i.e., to learn the smooth dependence of solutions on geometric deformations. Specifically, given a reference solution, RNO can predict solutions corresponding to arbitrary deformations of the referred geometry. This approach turns out to be much more data efficient. Through extensive experiments, we show that RNO can learn the dependence across various types and different numbers of geometry objects with relatively small datasets. RNO outperforms baseline models in accuracy by a large lead and achieves up to 80% error reduction. |
https://proceedings.mlr.press/v235/cheng24d.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/cheng24d/cheng24d.pdf | https://openreview.net/forum?id=Bb8pOvWIe4 | Causal Inference out of Control: Estimating Performativity without Treatment Randomization | https://proceedings.mlr.press/v235/cheng24d.html | Gary Cheng, Moritz Hardt, Celestine Mendler-Dünner | https://proceedings.mlr.press/v235/cheng24d.html | ICML 2024 | Regulators and academics are increasingly interested in the causal effect that algorithmic actions of a digital platform have on user consumption. In pursuit of estimating this effect from observational data, we identify a set of assumptions that permit causal identifiability without assuming randomized platform actions. Our results are applicable to platforms that rely on machine-learning-powered predictions and leverage knowledge from historical data. The key novelty of our approach is to explicitly model the dynamics of consumption over time, exploiting the repeated interaction of digital platforms with their participants to prove our identifiability results. By viewing the platform as a controller acting on a dynamical system, we can show that exogenous variation in consumption and appropriately responsive algorithmic control actions are sufficient for identifying the causal effect of interest. We complement our claims with an analysis of ready-to-use finite sample estimators and empirical investigations. More broadly, our results deriving identifiability conditions tailored to digital platform settings illustrate a fruitful interplay of control theory and causal inference. |
https://proceedings.mlr.press/v235/cheng24e.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/cheng24e/cheng24e.pdf | https://openreview.net/forum?id=uGoi3nY62g | BadPart: Unified Black-box Adversarial Patch Attacks against Pixel-wise Regression Tasks | https://proceedings.mlr.press/v235/cheng24e.html | Zhiyuan Cheng, Zhaoyi Liu, Tengda Guo, Shiwei Feng, Dongfang Liu, Mingjie Tang, Xiangyu Zhang | https://proceedings.mlr.press/v235/cheng24e.html | ICML 2024 | Pixel-wise regression tasks (e.g., monocular depth estimation (MDE) and optical flow estimation (OFE)) have been widely involved in our daily life in applications like autonomous driving, augmented reality and video composition. Although certain applications are security-critical or bear societal significance, the adversarial robustness of such models are not sufficiently studied, especially in the black-box scenario. In this work, we introduce the first unified black-box adversarial patch attack framework against pixel-wise regression tasks, aiming to identify the vulnerabilities of these models under query-based black-box attacks. We propose a novel square-based adversarial patch optimization framework and employ probabilistic square sampling and score-based gradient estimation techniques to generate the patch effectively and efficiently, overcoming the scalability problem of previous black-box patch attacks. Our attack prototype, named BadPart, is evaluated on both MDE and OFE tasks, utilizing a total of 7 models. BadPart surpasses 3 baseline methods in terms of both attack performance and efficiency. We also apply BadPart on the Google online service for portrait depth estimation, causing 43.5% relative distance error with 50K queries. State-of-the-art (SOTA) countermeasures cannot defend our attack effectively. |
https://proceedings.mlr.press/v235/cheng24f.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/cheng24f/cheng24f.pdf | https://openreview.net/forum?id=lQ3SEBH1gF | GaussianPro: 3D Gaussian Splatting with Progressive Propagation | https://proceedings.mlr.press/v235/cheng24f.html | Kai Cheng, Xiaoxiao Long, Kaizhi Yang, Yao Yao, Wei Yin, Yuexin Ma, Wenping Wang, Xuejin Chen | https://proceedings.mlr.press/v235/cheng24f.html | ICML 2024 | 3D Gaussian Splatting (3DGS) has recently revolutionized the field of neural rendering with its high fidelity and efficiency. However, 3DGS heavily depends on the initialized point cloud produced by Structure-from-Motion (SfM) techniques. When tackling large-scale scenes that unavoidably contain texture-less surfaces, SfM techniques fail to produce enough points in these surfaces and cannot provide good initialization for 3DGS. As a result, 3DGS suffers from difficult optimization and low-quality renderings. In this paper, inspired by classic multi-view stereo (MVS) techniques, we propose GaussianPro, a novel method that applies a progressive propagation strategy to guide the densification of the 3D Gaussians. Compared to the simple split and clone strategies used in 3DGS, our method leverages the priors of the existing reconstructed geometries of the scene and utilizes patch matching to produce new Gaussians with accurate positions and orientations. Experiments on both large-scale and small-scale scenes validate the effectiveness of our method. Our method significantly surpasses 3DGS on the Waymo dataset, exhibiting an improvement of 1.15dB in terms of PSNR. Codes and data are available at https://github.com/kcheng1021/GaussianPro. |
https://proceedings.mlr.press/v235/cheng24g.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/cheng24g/cheng24g.pdf | https://openreview.net/forum?id=DRBgNQ2N7U | Characterizing Overfitting in Kernel Ridgeless Regression Through the Eigenspectrum | https://proceedings.mlr.press/v235/cheng24g.html | Tin Sum Cheng, Aurelien Lucchi, Anastasis Kratsios, David Belius | https://proceedings.mlr.press/v235/cheng24g.html | ICML 2024 | We derive new bounds for the condition number of kernel matrices, which we then use to enhance existing non-asymptotic test error bounds for kernel ridgeless regression in the over-parameterized regime for a fixed input dimension. For kernels with polynomial spectral decay, we recover the bound from previous work; for exponential decay, our bound is non-trivial and novel. Our conclusion is two-fold: (i) kernel regressors whose eigenspectrum decays polynomially must generalize well, even in the presence of noisy labeled training data; these models exhibit so-called tempered overfitting; (ii) if the eigenspectrum of any kernel ridge regressor decays exponentially, then it generalizes poorly, i.e., it exhibits catastrophic overfitting. This adds to the available characterization of kernel ridge regressors exhibiting benign overfitting as the extremal case where the eigenspectrum of the kernel decays sub-polynomially. Our analysis combines new random matrix theory (RMT) techniques with recent tools in the kernel ridge regression (KRR) literature. |
https://proceedings.mlr.press/v235/cheng24h.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/cheng24h/cheng24h.pdf | https://openreview.net/forum?id=CR6Sl80cn8 | Efficient Black-box Adversarial Attacks via Bayesian Optimization Guided by a Function Prior | https://proceedings.mlr.press/v235/cheng24h.html | Shuyu Cheng, Yibo Miao, Yinpeng Dong, Xiao Yang, Xiao-Shan Gao, Jun Zhu | https://proceedings.mlr.press/v235/cheng24h.html | ICML 2024 | This paper studies the challenging black-box adversarial attack that aims to generate adversarial examples against a black-box model by only using output feedback of the model to input queries. Some previous methods improve the query efficiency by incorporating the gradient of a surrogate white-box model into query-based attacks due to the adversarial transferability. However, the localized gradient is not informative enough, making these methods still query-intensive. In this paper, we propose a Prior-guided Bayesian Optimization (P-BO) algorithm that leverages the surrogate model as a global function prior in black-box adversarial attacks. As the surrogate model contains rich prior information of the black-box one, P-BO models the attack objective with a Gaussian process whose mean function is initialized as the surrogate model’s loss. Our theoretical analysis on the regret bound indicates that the performance of P-BO may be affected by a bad prior. Therefore, we further propose an adaptive integration strategy to automatically adjust a coefficient on the function prior by minimizing the regret bound. Extensive experiments on image classifiers and large vision-language models demonstrate the superiority of the proposed algorithm in reducing queries and improving attack success rates compared with the state-of-the-art black-box attacks. Code is available at https://github.com/yibo-miao/PBO-Attack. |
https://proceedings.mlr.press/v235/cheng24i.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/cheng24i/cheng24i.pdf | https://openreview.net/forum?id=girxGkdECL | Can AI Assistants Know What They Don’t Know? | https://proceedings.mlr.press/v235/cheng24i.html | Qinyuan Cheng, Tianxiang Sun, Xiangyang Liu, Wenwei Zhang, Zhangyue Yin, Shimin Li, Linyang Li, Zhengfu He, Kai Chen, Xipeng Qiu | https://proceedings.mlr.press/v235/cheng24i.html | ICML 2024 | AI assistants powered by Large Language Models (LLMs) have demonstrated impressive performance in various tasks. However, LLMs still make factual errors in knowledge-intensive tasks such as open-domain question answering. These untruthful responses from AI assistants can pose significant risks in practical applications. Therefore, in this paper, we ask the question Can AI assistants know what they don’t know and express this awareness through natural language? To investigate this, we construct a model-specific "I don’t know" (Idk) dataset. This dataset includes Supervised Fine-tuning data and preference data, categorizing questions based on whether the assistant knows or does not know the answers. Then, we align the assistant with its corresponding Idk dataset using different alignment methods, including Supervised Fine-tuning and preference optimization. Experimental results show that, after alignment with the Idk dataset, the assistant is more capable of declining to answer questions outside its knowledge scope. The assistant aligned with the Idk dataset shows significantly higher truthfulness than the original assistant. |
https://proceedings.mlr.press/v235/cheng24j.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/cheng24j/cheng24j.pdf | https://openreview.net/forum?id=PKJqsZD5nQ | RICE: Breaking Through the Training Bottlenecks of Reinforcement Learning with Explanation | https://proceedings.mlr.press/v235/cheng24j.html | Zelei Cheng, Xian Wu, Jiahao Yu, Sabrina Yang, Gang Wang, Xinyu Xing | https://proceedings.mlr.press/v235/cheng24j.html | ICML 2024 | Deep reinforcement learning (DRL) is playing an increasingly important role in real-world applications. However, obtaining an optimally performing DRL agent for complex tasks, especially with sparse rewards, remains a significant challenge. The training of a DRL agent can be often trapped in a bottleneck without further progress. In this paper, we propose RICE, an innovative refining scheme for reinforcement learning that incorporates explanation methods to break through the training bottlenecks. The high-level idea of RICE is to construct a new initial state distribution that combines both the default initial states and critical states identified through explanation methods, thereby encouraging the agent to explore from the mixed initial states. Through careful design, we can theoretically guarantee that our refining scheme has a tighter sub-optimality bound. We evaluate RICE in various popular RL environments and real-world applications. The results demonstrate that RICE significantly outperforms existing refining schemes in enhancing agent performance. |
https://proceedings.mlr.press/v235/cheng24k.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/cheng24k/cheng24k.pdf | https://openreview.net/forum?id=BxAvcnlS8O | RIME: Robust Preference-based Reinforcement Learning with Noisy Preferences | https://proceedings.mlr.press/v235/cheng24k.html | Jie Cheng, Gang Xiong, Xingyuan Dai, Qinghai Miao, Yisheng Lv, Fei-Yue Wang | https://proceedings.mlr.press/v235/cheng24k.html | ICML 2024 | Preference-based Reinforcement Learning (PbRL) circumvents the need for reward engineering by harnessing human preferences as the reward signal. However, current PbRL methods excessively depend on high-quality feedback from domain experts, which results in a lack of robustness. In this paper, we present RIME, a robust PbRL algorithm for effective reward learning from noisy preferences. Our method utilizes a sample selection-based discriminator to dynamically filter out noise and ensure robust training. To counteract the cumulative error stemming from incorrect selection, we suggest a warm start for the reward model, which additionally bridges the performance gap during the transition from pre-training to online training in PbRL. Our experiments on robotic manipulation and locomotion tasks demonstrate that RIME significantly enhances the robustness of the state-of-the-art PbRL method. Code is available at https://github.com/CJReinforce/RIME_ICML2024. |
https://proceedings.mlr.press/v235/cheng24l.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/cheng24l/cheng24l.pdf | https://openreview.net/forum?id=w5oUo0LhO1 | Kernel Semi-Implicit Variational Inference | https://proceedings.mlr.press/v235/cheng24l.html | Ziheng Cheng, Longlin Yu, Tianyu Xie, Shiyue Zhang, Cheng Zhang | https://proceedings.mlr.press/v235/cheng24l.html | ICML 2024 | Semi-implicit variational inference (SIVI) extends traditional variational families with semi-implicit distributions defined in a hierarchical manner. Due to the intractable densities of semi-implicit distributions, classical SIVI often resorts to surrogates of evidence lower bound (ELBO) that would introduce biases for training. A recent advancement in SIVI, named SIVI-SM, utilizes an alternative score matching objective made tractable via a minimax formulation, albeit requiring an additional lower-level optimization. In this paper, we propose kernel SIVI (KSIVI), a variant of SIVI-SM that eliminates the need for the lower-level optimization through kernel tricks. Specifically, we show that when optimizing over a reproducing kernel Hilbert space (RKHS), the lower-level problem has an explicit solution. This way, the upper-level objective becomes the kernel Stein discrepancy (KSD), which is readily computable for stochastic gradient descent due to the hierarchical structure of semi-implicit variational distributions. An upper bound for the variance of the Monte Carlo gradient estimators of the KSD objective is derived, which allows us to establish novel convergence guarantees of KSIVI. We demonstrate the effectiveness and efficiency of KSIVI on both synthetic distributions and a variety of real data Bayesian inference tasks. |
https://proceedings.mlr.press/v235/cherep24a.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/cherep24a/cherep24a.pdf | https://openreview.net/forum?id=5pg9YJBaiG | Creative Text-to-Audio Generation via Synthesizer Programming | https://proceedings.mlr.press/v235/cherep24a.html | Manuel Cherep, Nikhil Singh, Jessica Shand | https://proceedings.mlr.press/v235/cherep24a.html | ICML 2024 | Neural audio synthesis methods now allow specifying ideas in natural language. However, these methods produce results that cannot be easily tweaked, as they are based on large latent spaces and up to billions of uninterpretable parameters. We propose a text-to-audio generation method that leverages a virtual modular sound synthesizer with only 78 parameters. Synthesizers have long been used by skilled sound designers for media like music and film due to their flexibility and intuitive controls. Our method, CTAG, iteratively updates a synthesizer’s parameters to produce high-quality audio renderings of text prompts that can be easily inspected and tweaked. Sounds produced this way are also more abstract, capturing essential conceptual features over fine-grained acoustic details, akin to how simple sketches can vividly convey visual concepts. Our results show how CTAG produces sounds that are distinctive, perceived as artistic, and yet similarly identifiable to recent neural audio synthesis models, positioning it as a valuable and complementary tool. |
https://proceedings.mlr.press/v235/cheung24a.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/cheung24a/cheung24a.pdf | https://openreview.net/forum?id=WvIHbQhrTq | Leveraging (Biased) Information: Multi-armed Bandits with Offline Data | https://proceedings.mlr.press/v235/cheung24a.html | Wang Chi Cheung, Lixing Lyu | https://proceedings.mlr.press/v235/cheung24a.html | ICML 2024 | We leverage offline data to facilitate online learning in stochastic multi-armed bandits. The probability distributions that govern the offline data and the online rewards can be different. Without any non-trival upper bound on their difference, we show that no non-anticipatory policy can out-perform the UCB policy by (Auer et al. 2002), even in the presence of offline data. In complement, we propose an online policy MIN-UCB, which outperforms UCB when a non-trivial upper bound is given. MIN-UCB adaptively chooses to utilize the offline data when they are deemed informative, and to ignore them otherwise. MIN-UCB is shown to be tight in terms of both instance indepedent and dependent regret bounds. Finally, we corroborate the theoretical results with numerical experiments. |
https://proceedings.mlr.press/v235/chevalier24a.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chevalier24a/chevalier24a.pdf | https://openreview.net/forum?id=WFyolnFZOR | Language Models as Science Tutors | https://proceedings.mlr.press/v235/chevalier24a.html | Alexis Chevalier, Jiayi Geng, Alexander Wettig, Howard Chen, Sebastian Mizera, Toni Annala, Max Aragon, Arturo Rodriguez Fanlo, Simon Frieder, Simon Machado, Akshara Prabhakar, Ellie Thieu, Jiachen T. Wang, Zirui Wang, Xindi Wu, Mengzhou Xia, Wenhan Xia, Jiatong Yu, Junjie Zhu, Zhiyong Ren, Sanjeev Arora, Danqi Chen | https://proceedings.mlr.press/v235/chevalier24a.html | ICML 2024 | NLP has recently made exciting progress toward training language models (LMs) with strong scientific problem-solving skills. However, model development has not focused on real-life use-cases of LMs for science, including applications in education that require processing long scientific documents. To address this, we introduce TutorEval and TutorChat. TutorEval is a diverse question-answering benchmark consisting of questions about long chapters from STEM textbooks, written by experts. TutorEval helps measure real-life usability of LMs as scientific assistants, and it is the first benchmark combining long contexts, free-form generation, and multi-disciplinary scientific knowledge. Moreover, we show that fine-tuning base models with existing dialogue datasets leads to poor performance on TutorEval. Therefore, we create TutorChat, a dataset of 80,000 long synthetic dialogues about textbooks. We use TutorChat to fine-tune Llemma models with 7B and 34B parameters. These LM tutors specialized in math have a 32K-token context window, and they excel at TutorEval while performing strongly on GSM8K and MATH. Our datasets build on open-source materials, and we release our models, data, and evaluations publicly. |
https://proceedings.mlr.press/v235/chiang24a.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chiang24a/chiang24a.pdf | https://openreview.net/forum?id=gzis9n5r7e | Expert Proximity as Surrogate Rewards for Single Demonstration Imitation Learning | https://proceedings.mlr.press/v235/chiang24a.html | Chia-Cheng Chiang, Li-Cheng Lan, Wei-Fang Sun, Chien Feng, Cho-Jui Hsieh, Chun-Yi Lee | https://proceedings.mlr.press/v235/chiang24a.html | ICML 2024 | In this paper, we focus on single-demonstration imitation learning (IL), a practical approach for real-world applications where acquiring multiple expert demonstrations is costly or infeasible and the ground truth reward function is not available. In contrast to typical IL settings with multiple demonstrations, single-demonstration IL involves an agent having access to only one expert trajectory. We highlight the issue of sparse reward signals in this setting and propose to mitigate this issue through our proposed Transition Discriminator-based IL (TDIL) method. TDIL is an IRL method designed to address reward sparsity by introducing a denser surrogate reward function that considers environmental dynamics. This surrogate reward function encourages the agent to navigate towards states that are proximal to expert states. In practice, TDIL trains a transition discriminator to differentiate between valid and non-valid transitions in a given environment to compute the surrogate rewards. The experiments demonstrate that TDIL outperforms existing IL approaches and achieves expert-level performance in the single-demonstration IL setting across five widely adopted MuJoCo benchmarks as well as the "Adroit Door" robotic environment. |
https://proceedings.mlr.press/v235/chiang24b.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chiang24b/chiang24b.pdf | https://openreview.net/forum?id=3MW8GKNyzI | Chatbot Arena: An Open Platform for Evaluating LLMs by Human Preference | https://proceedings.mlr.press/v235/chiang24b.html | Wei-Lin Chiang, Lianmin Zheng, Ying Sheng, Anastasios Nikolas Angelopoulos, Tianle Li, Dacheng Li, Banghua Zhu, Hao Zhang, Michael Jordan, Joseph E. Gonzalez, Ion Stoica | https://proceedings.mlr.press/v235/chiang24b.html | ICML 2024 | Large Language Models (LLMs) have unlocked new capabilities and applications; however, evaluating the alignment with human preferences still poses significant challenges. To address this issue, we introduce Chatbot Arena, an open platform for evaluating LLMs based on human preferences. Our methodology employs a pairwise comparison approach and leverages input from a diverse user base through crowdsourcing. The platform has been operational for several months, amassing over 240K votes. This paper describes the platform, analyzes the data we have collected so far, and explains the tried-and-true statistical methods we are using for efficient and accurate evaluation and ranking of models. We confirm that the crowdsourced questions are sufficiently diverse and discriminating and that the crowd-sourced human votes are in good agreement with those of expert raters. These analyses collectively establish a robust foundation for the credibility of Chatbot Arena. Because of its unique value and openness, Chatbot Arena has emerged as one of the most referenced LLM leaderboards, widely cited by leading LLM developers and companies. The platform is publicly available at https://chat.lmsys.org. |
https://proceedings.mlr.press/v235/chib24a.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chib24a/chib24a.pdf | https://openreview.net/forum?id=s4Hy0L4mml | MS-TIP: Imputation Aware Pedestrian Trajectory Prediction | https://proceedings.mlr.press/v235/chib24a.html | Pranav Singh Chib, Achintya Nath, Paritosh Kabra, Ishu Gupta, Pravendra Singh | https://proceedings.mlr.press/v235/chib24a.html | ICML 2024 | Pedestrian trajectory prediction aims to predict future trajectories based on observed trajectories. Current state-of-the-art methods often assume that the observed sequences of agents are complete, which is a strong assumption that overlooks inherent uncertainties. Understanding pedestrian behavior when dealing with missing values in the observed sequence is crucial for enhancing the performance of predictive models. In this work, we propose the MultiScale hypergraph for Trajectory Imputation and Prediction (MS-TIP), a novel approach that simultaneously addresses the imputation of missing observations and the prediction of future trajectories. Specifically, we leverage transformers with diagonal masked self-attention to impute incomplete observations. Further, our approach promotes complex interaction modeling through multi-scale hypergraphs, optimizing our trajectory prediction module to capture different types of interactions. With the inclusion of scenic attention, we learn contextual scene information, instead of sole reliance on coordinates. Additionally, our approach utilizes an intermediate control point and refinement module to infer future trajectories accurately. Extensive experiments validate the efficacy of MS-TIP in precisely predicting pedestrian future trajectories. Code is publicly available at https://github.com/Pranav-chib/MS-TIP. |
https://proceedings.mlr.press/v235/chib24b.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chib24b/chib24b.pdf | https://openreview.net/forum?id=OQ7TlOphGX | Enhancing Trajectory Prediction through Self-Supervised Waypoint Distortion Prediction | https://proceedings.mlr.press/v235/chib24b.html | Pranav Singh Chib, Pravendra Singh | https://proceedings.mlr.press/v235/chib24b.html | ICML 2024 | Trajectory prediction is an important task that involves modeling the indeterminate nature of agents to forecast future trajectories given the observed trajectory sequences. The task of predicting trajectories poses significant challenges, as agents not only move individually through time but also interact spatially. The learning of complex spatio-temporal representations stands as a fundamental challenge in trajectory prediction. To this end, we propose a novel approach called SSWDP (Self-Supervised Waypoint Distortion Prediction). We propose a simple yet highly effective self-supervised task of predicting distortion present in the observed trajectories to improve the representation learning of the model. Our approach can complement existing trajectory prediction methods. The experimental results highlight a significant improvement with relative percentage differences of 22.7%/38.9%, 33.8%/36.4%, and 16.60%/23.20% in ADE/FDE for the NBA, TrajNet++, and ETH-UCY datasets, respectively, compared to the baseline methods. Our approach also demonstrates a significant improvement over baseline methods with relative percentage differences of 76.8%/82.5% and 61.0%/36.1% in ADE/FDE for TrajNet++ and NBA datasets in distorted environments, respectively. |
https://proceedings.mlr.press/v235/chidambaram24a.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chidambaram24a/chidambaram24a.pdf | https://openreview.net/forum?id=3McL91pE6x | How Flawed Is ECE? An Analysis via Logit Smoothing | https://proceedings.mlr.press/v235/chidambaram24a.html | Muthu Chidambaram, Holden Lee, Colin Mcswiggen, Semon Rezchikov | https://proceedings.mlr.press/v235/chidambaram24a.html | ICML 2024 | Informally, a model is calibrated if its predictions are correct with a probability that matches the confidence of the prediction. By far the most common method in the literature for measuring calibration is the expected calibration error (ECE). Recent work, however, has pointed out drawbacks of ECE, such as the fact that it is discontinuous in the space of predictors. In this work, we ask: how fundamental are these issues, and what are their impacts on existing results? Towards this end, we completely characterize the discontinuities of ECE with respect to general probability measures on Polish spaces. We then use the nature of these discontinuities to motivate a novel continuous, easily estimated miscalibration metric, which we term Logit-Smoothed ECE (LS-ECE). By comparing the ECE and LS-ECE of pre-trained image classification models, we show in initial experiments that binned ECE closely tracks LS-ECE, indicating that the theoretical pathologies of ECE may be avoidable in practice. |
https://proceedings.mlr.press/v235/chien24a.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chien24a/chien24a.pdf | https://openreview.net/forum?id=ZUXvpIrz5l | Safe Exploration in Dose Finding Clinical Trials with Heterogeneous Participants | https://proceedings.mlr.press/v235/chien24a.html | Isabel Chien, Wessel P Bruinsma, Javier Gonzalez, Richard E. Turner | https://proceedings.mlr.press/v235/chien24a.html | ICML 2024 | In drug development, early phase dose-finding clinical trials are carried out to identify an optimal dose to administer to patients in larger confirmatory clinical trials. Standard trial procedures do not optimize for participant benefit and do not consider participant heterogeneity, despite consequences to participants’ health and downstream impacts to under-represented population subgroups. Many novel drugs also do not obey parametric modelling assumptions made in common dose-finding procedures. We present Safe Allocation for Exploration of Treatments SAFE-T, a procedure for adaptive dose-finding that adheres to safety constraints, improves utility for heterogeneous participants, and works well with small sample sizes. SAFE-T flexibly learns non-parametric multi-output Gaussian process models for dose toxicity and efficacy, using Bayesian optimization, and provides accurate final dose recommendations. We provide theoretical guarantees for the satisfaction of safety constraints. Using a comprehensive set of realistic synthetic scenarios, we demonstrate empirically that SAFE-T generally outperforms comparable methods and maintains performance across variations in sample size and subgroup distribution. Finally, we extend SAFE-T to a new adaptive setting, demonstrating its potential to improve traditional clinical trial procedures. |
https://proceedings.mlr.press/v235/chin24a.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chin24a/chin24a.pdf | https://openreview.net/forum?id=VyGo1S5A6d | Prompting4Debugging: Red-Teaming Text-to-Image Diffusion Models by Finding Problematic Prompts | https://proceedings.mlr.press/v235/chin24a.html | Zhi-Yi Chin, Chieh Ming Jiang, Ching-Chun Huang, Pin-Yu Chen, Wei-Chen Chiu | https://proceedings.mlr.press/v235/chin24a.html | ICML 2024 | Text-to-image diffusion models, e.g. Stable Diffusion (SD), lately have shown remarkable ability in high-quality content generation, and become one of the representatives for the recent wave of transformative AI. Nevertheless, such advance comes with an intensifying concern about the misuse of this generative technology, especially for producing copyrighted or NSFW (i.e. not safe for work) images. Although efforts have been made to filter inappropriate images/prompts or remove undesirable concepts/styles via model fine-tuning, the reliability of these safety mechanisms against diversified problematic prompts remains largely unexplored. In this work, we propose Prompting4Debugging (P4D) as a debugging and red-teaming tool that automatically finds problematic prompts for diffusion models to test the reliability of a deployed safety mechanism. We demonstrate the efficacy of our P4D tool in uncovering new vulnerabilities of SD models with safety mechanisms. Particularly, our result shows that around half of prompts in existing safe prompting benchmarks which were originally considered "safe" can actually be manipulated to bypass many deployed safety mechanisms, including concept removal, negative prompt, and safety guidance. Our findings suggest that, without comprehensive testing, the evaluations on limited safe prompting benchmarks can lead to a false sense of safety for text-to-image models. |
https://proceedings.mlr.press/v235/cho24a.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/cho24a/cho24a.pdf | https://openreview.net/forum?id=hcASxFvmZ5 | Peeking with PEAK: Sequential, Nonparametric Composite Hypothesis Tests for Means of Multiple Data Streams | https://proceedings.mlr.press/v235/cho24a.html | Brian M Cho, Kyra Gan, Nathan Kallus | https://proceedings.mlr.press/v235/cho24a.html | ICML 2024 | We propose a novel nonparametric sequential test for composite hypotheses for means of multiple data streams. Our proposed method, peeking with expectation-based averaged capital (PEAK), builds upon the testing-by-betting framework and provides a non-asymptotic $\alpha$-level test across any stopping time. Our contributions are two-fold: (1) we propose a novel betting scheme and provide theoretical guarantees on type-I error control, power, and asymptotic growth rate/$e$-power in the setting of a single data stream; (2) we introduce PEAK, a generalization of this betting scheme to multiple streams, that (i) avoids using wasteful union bounds via averaging, (ii) is a test of power one under mild regularity conditions on the sampling scheme of the streams, and (iii) reduces computational overhead when applying the testing-as-betting approaches for pure-exploration bandit problems. We illustrate the practical benefits of PEAK using both synthetic and real-world HeartSteps datasets. Our experiments show that PEAK provides up to an 85% reduction in the number of samples before stopping compared to existing stopping rules for pure-exploration bandit problems, and matches the performance of state-of-the-art sequential tests while improving upon computational complexity. |
https://proceedings.mlr.press/v235/cho24b.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/cho24b/cho24b.pdf | https://openreview.net/forum?id=n3yYrtt9U7 | Parameterized Physics-informed Neural Networks for Parameterized PDEs | https://proceedings.mlr.press/v235/cho24b.html | Woojin Cho, Minju Jo, Haksoo Lim, Kookjin Lee, Dongeun Lee, Sanghyun Hong, Noseong Park | https://proceedings.mlr.press/v235/cho24b.html | ICML 2024 | Complex physical systems are often described by partial differential equations (PDEs) that depend on parameters such as the Raynolds number in fluid mechanics. In applications such as design optimization or uncertainty quantification, solutions of those PDEs need to be evaluated at numerous points in the parameter space. While physics-informed neural networks (PINNs) have emerged as a new strong competitor as a surrogate, their usage in this scenario remains underexplored due to the inherent need for repetitive and time-consuming training. In this paper, we address this problem by proposing a novel extension, parameterized physics-informed neural networks (P$^2$INNs). P$^2$INNs enable modeling the solutions of parameterized PDEs via explicitly encoding a latent representation of PDE parameters. With the extensive empirical evaluation, we demonstrate that P$^2$INNs outperform the baselines both in accuracy and parameter efficiency on benchmark 1D and 2D parameterized PDEs and are also effective in overcoming the known “failure modes”. |
https://proceedings.mlr.press/v235/cho24c.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/cho24c/cho24c.pdf | https://openreview.net/forum?id=vq7ITv8a49 | Kernel Debiased Plug-in Estimation: Simultaneous, Automated Debiasing without Influence Functions for Many Target Parameters | https://proceedings.mlr.press/v235/cho24c.html | Brian M Cho, Yaroslav Mukhin, Kyra Gan, Ivana Malenica | https://proceedings.mlr.press/v235/cho24c.html | ICML 2024 | When estimating target parameters in nonparametric models with nuisance parameters, substituting the unknown nuisances with nonparametric estimators can introduce "plug-in bias." Traditional methods addressing this suboptimal bias-variance trade-off rely on the influence function (IF) of the target parameter. When estimating multiple target parameters, these methods require debiasing the nuisance parameter multiple times using the corresponding IFs, which poses analytical and computational challenges. In this work, we leverage the targeted maximum likelihood estimation (TMLE) framework to propose a novel method named kernel debiased plug-in estimation (KDPE). KDPE refines an initial estimate through regularized likelihood maximization steps, employing a nonparametric model based on reproducing kernel Hilbert spaces. We show that KDPE: (i) simultaneously debiases all pathwise differentiable target parameters that satisfy our regularity conditions, (ii) does not require the IF for implementation, and (iii) remains computationally tractable. We numerically illustrate the use of KDPE and validate our theoretical results. |
https://proceedings.mlr.press/v235/cho24d.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/cho24d/cho24d.pdf | https://openreview.net/forum?id=haUOhXo70o | Hard Tasks First: Multi-Task Reinforcement Learning Through Task Scheduling | https://proceedings.mlr.press/v235/cho24d.html | Myungsik Cho, Jongeui Park, Suyoung Lee, Youngchul Sung | https://proceedings.mlr.press/v235/cho24d.html | ICML 2024 | Multi-task reinforcement learning (RL) faces the significant challenge of varying task difficulties, often leading to negative transfer when simpler tasks overshadow the learning of more complex ones. To overcome this challenge, we propose a novel algorithm, Scheduled Multi-Task Training (SMT), that strategically prioritizes more challenging tasks, thereby enhancing overall learning efficiency. SMT introduces a dynamic task prioritization strategy, underpinned by an effective metric for assessing task difficulty. This metric ensures an efficient and targeted allocation of training resources, significantly improving learning outcomes. Additionally, SMT incorporates a reset mechanism that periodically reinitializes key network parameters to mitigate the simplicity bias, further enhancing the adaptability and robustness of the learning process across diverse tasks. The efficacy of SMT’s scheduling method is validated by significantly improving performance on challenging Meta-World benchmarks. |
https://proceedings.mlr.press/v235/cho24e.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/cho24e/cho24e.pdf | https://openreview.net/forum?id=OBs0AjXE3F | KV-Runahead: Scalable Causal LLM Inference by Parallel Key-Value Cache Generation | https://proceedings.mlr.press/v235/cho24e.html | Minsik Cho, Mohammad Rastegari, Devang Naik | https://proceedings.mlr.press/v235/cho24e.html | ICML 2024 | Large Language Model or LLM inference has two phases, the prompt (or prefill) phase to output the first token and the extension (or decoding) phase to the generate subsequent tokens. In this work, we propose an efficient parallelization scheme, KV-Runahead to accelerate the prompt phase. The key observation is that the extension phase generates tokens faster than the prompt phase because of key-value cache (KV-cache). Hence, KV-Runahead parallelizes the prompt phase by orchestrating multiple processes to populate the KV-cache and minimizes the time-to-first-token (TTFT). Dual-purposing the KV-cache scheme has two main benefits. First, since KV-cache is designed to leverage the causal attention map, we minimize computation and computation automatically. Second, since it already exists for the extension phase, KV-Runahead is easy to implement. We further propose context-level load-balancing to handle uneven KV-cache generation (due to the causal attention) and to optimize TTFT. Compared with an existing parallelization scheme such as tensor or sequential parallelization where keys and values are locally generated and exchanged via all-gather collectives, our experimental results demonstrate that KV-Runahead can offer over 1.4× and 1.6× speedups for Llama 7B and Falcon 7B respectively. |
https://proceedings.mlr.press/v235/cho24f.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/cho24f/cho24f.pdf | https://openreview.net/forum?id=GTnn6bNE3j | Neurodegenerative Brain Network Classification via Adaptive Diffusion with Temporal Regularization | https://proceedings.mlr.press/v235/cho24f.html | Hyuna Cho, Jaeyoon Sim, Guorong Wu, Won Hwa Kim | https://proceedings.mlr.press/v235/cho24f.html | ICML 2024 | Analysis of neurodegenerative diseases on brain connectomes is important in facilitating early diagnosis and predicting its onset. However, investigation of the progressive and irreversible dynamics of these diseases remains underexplored in cross-sectional studies as its diagnostic groups are considered independent. Also, as in many real-world graphs, brain networks exhibit intricate structures with both homophily and heterophily. To address these challenges, we propose Adaptive Graph diffusion network with Temporal regularization (AGT). AGT introduces node-wise convolution to adaptively capture low (i.e., homophily) and high-frequency (i.e., heterophily) characteristics within an optimally tailored range for each node. Moreover, AGT captures sequential variations within progressive diagnostic groups with a novel temporal regularization, considering the relative feature distance between the groups in the latent space. As a result, our proposed model yields interpretable results at both node-level and group-level. The superiority of our method is validated on two neurodegenerative disease benchmarks for graph classification: Alzheimer’s Disease Neuroimaging Initiative (ADNI) and Parkinson’s Progression Markers Initiative (PPMI) datasets. |
https://proceedings.mlr.press/v235/cho24g.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/cho24g/cho24g.pdf | https://openreview.net/forum?id=61A1bsVjRg | Tilt and Average : Geometric Adjustment of the Last Layer for Recalibration | https://proceedings.mlr.press/v235/cho24g.html | Gyusang Cho, Chan-Hyun Youn | https://proceedings.mlr.press/v235/cho24g.html | ICML 2024 | After the revelation that neural networks tend to produce overconfident predictions, the problem of calibration, which aims to align confidence with accuracy to enhance the reliability of predictions, has gained significant importance. Several solutions based on calibration maps have been proposed to address the problem of recalibrating a trained classifier using additional datasets. In this paper, we offer an algorithm that transforms the weights of the last layer of the classifier, distinct from the calibration-map-based approach. We concentrate on the geometry of the final linear layer, specifically its angular aspect, and adjust the weights of the corresponding layer. We name the method Tilt and Average, and validate the calibration effect empirically and theoretically. Through this, we demonstrate that our approach, in addition to the existing calibration-map-based techniques, can yield improved calibration performance. |
https://proceedings.mlr.press/v235/choi24a.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/choi24a/choi24a.pdf | https://openreview.net/forum?id=dMhF96PfQi | Scalable Wasserstein Gradient Flow for Generative Modeling through Unbalanced Optimal Transport | https://proceedings.mlr.press/v235/choi24a.html | Jaemoo Choi, Jaewoong Choi, Myungjoo Kang | https://proceedings.mlr.press/v235/choi24a.html | ICML 2024 | Wasserstein gradient flow (WGF) describes the gradient dynamics of probability density within the Wasserstein space. WGF provides a promising approach for conducting optimization over the probability distributions. Numerically approximating the continuous WGF requires the time discretization method. The most well-known method for this is the JKO scheme. In this regard, previous WGF models employ the JKO scheme and parametrized transport map for each JKO step. However, this approach results in quadratic training complexity $O(K^2)$ with the number of JKO step $K$. This severely limits the scalability of WGF models. In this paper, we introduce a scalable WGF-based generative model, called Semi-dual JKO (S-JKO). Our model is based on the semi-dual form of the JKO step, derived from the equivalence between the JKO step and the Unbalanced Optimal Transport. Our approach reduces the training complexity to $O(K)$. We demonstrate that our model significantly outperforms existing WGF-based generative models, achieving FID scores of 2.62 on CIFAR-10 and 6.42 on CelebA-HQ-256, which are comparable to state-of-the-art image generative models. |
https://proceedings.mlr.press/v235/choi24b.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/choi24b/choi24b.pdf | https://openreview.net/forum?id=If6Q9OYfoJ | Listwise Reward Estimation for Offline Preference-based Reinforcement Learning | https://proceedings.mlr.press/v235/choi24b.html | Heewoong Choi, Sangwon Jung, Hongjoon Ahn, Taesup Moon | https://proceedings.mlr.press/v235/choi24b.html | ICML 2024 | In Reinforcement Learning (RL), designing precise reward functions remains to be a challenge, particularly when aligning with human intent. Preference-based RL (PbRL) was introduced to address this problem by learning reward models from human feedback. However, existing PbRL methods have limitations as they often overlook the second-order preference that indicates the relative strength of preference. In this paper, we propose Listwise Reward Estimation (LiRE), a novel approach for offline PbRL that leverages second-order preference information by constructing a Ranked List of Trajectories (RLT), which can be efficiently built by using the same ternary feedback type as traditional methods. To validate the effectiveness of LiRE, we propose a new offline PbRL dataset that objectively reflects the effect of the estimated rewards. Our extensive experiments on the dataset demonstrate the superiority of LiRE, i.e., outperforming state-of-the-art baselines even with modest feedback budgets and enjoying robustness with respect to the number of feedbacks and feedback noise. Our code is available at https://github.com/chwoong/LiRE |
https://proceedings.mlr.press/v235/choi24c.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/choi24c/choi24c.pdf | https://openreview.net/forum?id=oWYzIodyC4 | BWS: Best Window Selection Based on Sample Scores for Data Pruning across Broad Ranges | https://proceedings.mlr.press/v235/choi24c.html | Hoyong Choi, Nohyun Ki, Hye Won Chung | https://proceedings.mlr.press/v235/choi24c.html | ICML 2024 | Data subset selection aims to find a smaller yet informative subset of a large dataset that can approximate the full-dataset training, addressing challenges associated with training neural networks on large-scale datasets. However, existing methods tend to specialize in either high or low selection ratio regimes, lacking a universal approach that consistently achieves competitive performance across a broad range of selection ratios. We introduce a universal and efficient data subset selection method, Best Window Selection (BWS), by proposing a method to choose the best window subset from samples ordered based on their difficulty scores. This approach offers flexibility by allowing the choice of window intervals that span from easy to difficult samples. Furthermore, we provide an efficient mechanism for selecting the best window subset by evaluating its quality using kernel ridge regression. Our experimental results demonstrate the superior performance of BWS compared to other baselines across a broad range of selection ratios over datasets, including CIFAR-10/100 and ImageNet, and the scenarios involving training from random initialization or fine-tuning of pre-trained models. |
https://proceedings.mlr.press/v235/choi24d.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/choi24d/choi24d.pdf | https://openreview.net/forum?id=M4Htd52HMH | Embodied CoT Distillation From LLM To Off-the-shelf Agents | https://proceedings.mlr.press/v235/choi24d.html | Wonje Choi, Woo Kyung Kim, Minjong Yoo, Honguk Woo | https://proceedings.mlr.press/v235/choi24d.html | ICML 2024 | We address the challenge of utilizing large language models (LLMs) for complex embodied tasks, in the environment where decision-making systems operate timely on capacity-limited, off-the-shelf devices. We present DeDer, a framework for decomposing and distilling the embodied reasoning capabilities from LLMs to efficient, small language model (sLM)-based policies. In DeDer, the decision-making process of LLM-based strategies is restructured into a hierarchy with a reasoning-policy and planning-policy. The reasoning-policy is distilled from the data that is generated through the embodied in-context learning and self-verification of an LLM, so it can produce effective rationales. The planning-policy, guided by the rationales, can render optimized plans efficiently. In turn, DeDer allows for adopting sLMs for both policies, deployed on off-the-shelf devices. Furthermore, to enhance the quality of intermediate rationales, specific to embodied tasks, we devise the embodied knowledge graph, and to generate multiple rationales timely through a single inference, we also use the contrastively prompted attention model. Our experiments with the ALFRED benchmark demonstrate that DeDer surpasses leading language planning and distillation approaches, indicating the applicability and efficiency of sLM-based embodied policies derived through DeDer. |
https://proceedings.mlr.press/v235/choi24e.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/choi24e/choi24e.pdf | https://openreview.net/forum?id=w1HdBXSJXn | PICLe: Eliciting Diverse Behaviors from Large Language Models with Persona In-Context Learning | https://proceedings.mlr.press/v235/choi24e.html | Hyeong Kyu Choi, Yixuan Li | https://proceedings.mlr.press/v235/choi24e.html | ICML 2024 | Large Language Models (LLMs) are trained on massive text corpora, which are encoded with diverse personality traits. This triggers an interesting goal of eliciting a desired personality trait from the LLM, and probing its behavioral preferences. Accordingly, we formalize the persona elicitation task, aiming to customize LLM behaviors to align with a target persona. We present Persona In-Context Learning (PICLe), a novel persona elicitation framework grounded in Bayesian inference. At the core, PICLe introduces a new ICL example selection criterion based on likelihood ratio, which is designed to optimally guide the model in eliciting a specific target persona. We demonstrate the effectiveness of PICLe through extensive comparisons against baseline methods across three contemporary LLMs. Code is available at https://github.com/deeplearning-wisc/picle. |
https://proceedings.mlr.press/v235/choi24f.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/choi24f/choi24f.pdf | https://openreview.net/forum?id=J1NIXxiDbu | PANDA: Expanded Width-Aware Message Passing Beyond Rewiring | https://proceedings.mlr.press/v235/choi24f.html | Jeongwhan Choi, Sumin Park, Hyowon Wi, Sung-Bae Cho, Noseong Park | https://proceedings.mlr.press/v235/choi24f.html | ICML 2024 | Recent research in the field of graph neural network (GNN) has identified a critical issue known as "over-squashing," resulting from the bottleneck phenomenon in graph structures, which impedes the propagation of long-range information. Prior works have proposed a variety of graph rewiring concepts that aim at optimizing the spatial or spectral properties of graphs to promote the signal propagation. However, such approaches inevitably deteriorate the original graph topology, which may lead to a distortion of information flow. To address this, we introduce an expanded width-aware (PANDA) message passing, a new message passing paradigm where nodes with high centrality, a potential source of over-squashing, are selectively expanded in width to encapsulate the growing influx of signals from distant nodes. Experimental results show that our method outperforms existing rewiring methods, suggesting that selectively expanding the hidden state of nodes can be a compelling alternative to graph rewiring for addressing the over-squashing. |
https://proceedings.mlr.press/v235/choo24a.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/choo24a/choo24a.pdf | https://openreview.net/forum?id=61WtHsVKWF | Online bipartite matching with imperfect advice | https://proceedings.mlr.press/v235/choo24a.html | Davin Choo, Themistoklis Gouleakis, Chun Kai Ling, Arnab Bhattacharyya | https://proceedings.mlr.press/v235/choo24a.html | ICML 2024 | We study the problem of online unweighted bipartite matching with $n$ offline vertices and $n$ online vertices where one wishes to be competitive against the optimal offline algorithm. While the classic RANKING algorithm of (Karp et al., 1990) provably attains competitive ratio of $1-1/e > 1/2$, we show that no learning-augmented method can be both 1-consistent and strictly better than 1/2-robust under the adversarial arrival model. Meanwhile, under the random arrival model, we show how one can utilize methods from distribution testing to design an algorithm that takes in external advice about the online vertices and provably achieves competitive ratio interpolating between any ratio attainable by advice-free methods and the optimal ratio of 1, depending on the advice quality. |
https://proceedings.mlr.press/v235/chopin24a.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chopin24a/chopin24a.pdf | https://openreview.net/forum?id=BtbijvkWLC | A connection between Tempering and Entropic Mirror Descent | https://proceedings.mlr.press/v235/chopin24a.html | Nicolas Chopin, Francesca Crucinio, Anna Korba | https://proceedings.mlr.press/v235/chopin24a.html | ICML 2024 | This paper explores the connections between tempering (for Sequential Monte Carlo; SMC) and entropic mirror descent to sample from a target probability distribution whose unnormalized density is known. We establish that tempering SMC corresponds to entropic mirror descent applied to the reverse Kullback-Leibler (KL) divergence and obtain convergence rates for the tempering iterates. Our result motivates the tempering iterates from an optimization point of view, showing that tempering can be seen as a descent scheme of the KL divergence with respect to the Fisher-Rao geometry, in contrast to Langevin dynamics that perform descent of the KL with respect to the Wasserstein-2 geometry. We exploit the connection between tempering and mirror descent iterates to justify common practices in SMC and derive adaptive tempering rules that improve over other alternative benchmarks in the literature. |
https://proceedings.mlr.press/v235/choukroun24a.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/choukroun24a/choukroun24a.pdf | https://openreview.net/forum?id=Kf9CqdI8Rb | Learning Linear Block Error Correction Codes | https://proceedings.mlr.press/v235/choukroun24a.html | Yoni Choukroun, Lior Wolf | https://proceedings.mlr.press/v235/choukroun24a.html | ICML 2024 | Error correction codes are a crucial part of the physical communication layer, ensuring the reliable transfer of data over noisy channels. The design of optimal linear block codes capable of being efficiently decoded is of major concern, especially for short block lengths. While neural decoders have recently demonstrated their advantage over classical decoding techniques, the neural design of the codes remains a challenge. In this work, we propose for the first time a unified encoder-decoder training of binary linear block codes. To this end, we adapt the coding setting to support efficient and differentiable training of the code for end-to-end optimization over the order two Galois field. We also propose a novel Transformer model in which the self-attention masking is performed in a differentiable fashion for the efficient backpropagation of the code gradient. Our results show that (i) the proposed decoder outperforms existing neural decoding on conventional codes, (ii) the suggested framework generates codes that outperform the analogous conventional codes, and (iii) the codes we developed not only excel with our decoder but also show enhanced performance with traditional decoding techniques. |
https://proceedings.mlr.press/v235/chowdhury24a.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chowdhury24a/chowdhury24a.pdf | https://openreview.net/forum?id=1oU4FKpVx5 | A Provably Effective Method for Pruning Experts in Fine-tuned Sparse Mixture-of-Experts | https://proceedings.mlr.press/v235/chowdhury24a.html | Mohammed Nowaz Rabbani Chowdhury, Meng Wang, Kaoutar El Maghraoui, Naigang Wang, Pin-Yu Chen, Christopher Carothers | https://proceedings.mlr.press/v235/chowdhury24a.html | ICML 2024 | The sparsely gated mixture of experts (MoE) architecture sends different inputs to different subnetworks (experts), through trainable routers. MoE reduces the training computation significantly for large models, but its deployment can be still memory/computation expensive for some downstream tasks. Model pruning is a popular approach to reduce inference computation, but its application in MoE architecture is largely unexplored. To the best of our knowledge, this paper provides the first provably efficient technique for pruning experts in fine-tuned MoE models. We theoretically prove that prioritizing the pruning of the experts with a smaller change of the router’s $l_2$ norm from the pre-trained model guarantees the preservation of test accuracy, while significantly reducing the model size and the computational requirements. Although our theoretical analysis is centered on binary classification tasks on simplified MoE architecture, our expert pruning method is verified on large vision MoE models such as V-MoE and $\text{E}^3$-MoE fine-tuned on benchmark datasets such as CIFAR-10, CIFAR-100, and ImageNet. |
https://proceedings.mlr.press/v235/chu24a.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chu24a/chu24a.pdf | https://openreview.net/forum?id=1YMjzz2g81 | SPABA: A Single-Loop and Probabilistic Stochastic Bilevel Algorithm Achieving Optimal Sample Complexity | https://proceedings.mlr.press/v235/chu24a.html | Tianshu Chu, Dachuan Xu, Wei Yao, Jin Zhang | https://proceedings.mlr.press/v235/chu24a.html | ICML 2024 | While stochastic bilevel optimization methods have been extensively studied for addressing large-scale nested optimization problems in machine learning, it remains an open question whether the optimal complexity bounds for solving bilevel optimization are the same as those in single-level optimization. Our main result resolves this question: SPABA, an adaptation of the PAGE method for nonconvex optimization in (Li et al., 2021) to the bilevel setting, can achieve optimal sample complexity in both the finite-sum and expectation settings. We show the optimality of SPABA by proving that there is no gap in complexity analysis between stochastic bilevel and single-level optimization when implementing PAGE. Notably, as indicated by the results of (Dagréou et al., 2022), there might exist a gap in complexity analysis when implementing other stochastic gradient estimators, like SGD and SAGA. In addition to SPABA, we propose several other single-loop stochastic bilevel algorithms, that either match or improve the state-of-the-art sample complexity results, leveraging our convergence rate and complexity analysis. Numerical experiments demonstrate the superior practical performance of the proposed methods. |
https://proceedings.mlr.press/v235/chua24a.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chua24a/chua24a.pdf | https://openreview.net/forum?id=xWI0MKwJSS | How Private are DP-SGD Implementations? | https://proceedings.mlr.press/v235/chua24a.html | Lynn Chua, Badih Ghazi, Pritish Kamath, Ravi Kumar, Pasin Manurangsi, Amer Sinha, Chiyuan Zhang | https://proceedings.mlr.press/v235/chua24a.html | ICML 2024 | We demonstrate a substantial gap between the privacy guarantees of the Adaptive Batch Linear Queries (ABLQ) mechanism under different types of batch sampling: (i) Shuffling, and (ii) Poisson subsampling; the typical analysis of Differentially Private Stochastic Gradient Descent (DP-SGD) follows by interpreting it as a post-processing of ABLQ. While shuffling-based DP-SGD is more commonly used in practical implementations, it has not been amenable to easy privacy analysis, either analytically or even numerically. On the other hand, Poisson subsampling-based DP-SGD is challenging to scalably implement, but has a well-understood privacy analysis, with multiple open-source numerically tight privacy accountants available. This has led to a common practice of using shuffling-based DP-SGD in practice, but using the privacy analysis for the corresponding Poisson subsampling version. Our result shows that there can be a substantial gap between the privacy analysis when using the two types of batch sampling, and thus advises caution in reporting privacy parameters for DP-SGD. |
https://proceedings.mlr.press/v235/chung24a.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chung24a/chung24a.pdf | https://openreview.net/forum?id=iJWeK2snMH | Sampling-based Multi-dimensional Recalibration | https://proceedings.mlr.press/v235/chung24a.html | Youngseog Chung, Ian Char, Jeff Schneider | https://proceedings.mlr.press/v235/chung24a.html | ICML 2024 | Calibration of probabilistic forecasts in the regression setting has been widely studied in the single dimensional case, where the output variables are assumed to be univariate. In many problem settings, however, the output variables are multi-dimensional, and in the presence of dependence across the output dimensions, measuring calibration and performing recalibration for each dimension separately can be both misleading and detrimental. In this work, we focus on representing predictive uncertainties via samples, and propose a recalibration method which accounts for the joint distribution across output dimensions to produce calibrated samples. Based on the concept of highest density regions (HDR), we define the notion of HDR calibration, and show that our recalibration method produces samples which are HDR calibrated. We demonstrate the performance of our method and the quality of the recalibrated samples on a suite of benchmark datasets in multi-dimensional regression, a real-world dataset in modeling plasma dynamics during nuclear fusion reactions, and on a decision-making application in forecasting demand. |
https://proceedings.mlr.press/v235/chung24b.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/chung24b/chung24b.pdf | https://openreview.net/forum?id=hrwIndai8e | Prompt-tuning Latent Diffusion Models for Inverse Problems | https://proceedings.mlr.press/v235/chung24b.html | Hyungjin Chung, Jong Chul Ye, Peyman Milanfar, Mauricio Delbracio | https://proceedings.mlr.press/v235/chung24b.html | ICML 2024 | We propose a new method for solving imaging inverse problems using text-to-image latent diffusion models as general priors. Existing methods using latent diffusion models for inverse problems typically rely on simple null text prompts, which can lead to suboptimal performance. To improve upon this, we introduce a method for prompt tuning, which jointly optimizes the text embedding on-the-fly while running the reverse diffusion. This allows us to generate images that are more faithful to the diffusion prior. Specifically, our approach involves a unified optimization framework that simultaneously considers the prompt, latent, and pixel values through alternating minimization. This significantly diminishes image artifacts - a major problem when using latent diffusion models instead of pixel-based diffusion ones. Our method, called P2L, outperforms both pixel- and latent-diffusion model-based inverse problem solvers on a variety of tasks, such as super-resolution, deblurring, and inpainting. Furthermore, P2L demonstrates remarkable scalability to higher resolutions without artifacts. |
https://proceedings.mlr.press/v235/cideron24a.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/cideron24a/cideron24a.pdf | https://openreview.net/forum?id=EruV94XRDs | MusicRL: Aligning Music Generation to Human Preferences | https://proceedings.mlr.press/v235/cideron24a.html | Geoffrey Cideron, Sertan Girgin, Mauro Verzetti, Damien Vincent, Matej Kastelic, Zalán Borsos, Brian Mcwilliams, Victor Ungureanu, Olivier Bachem, Olivier Pietquin, Matthieu Geist, Leonard Hussenot, Neil Zeghidour, Andrea Agostinelli | https://proceedings.mlr.press/v235/cideron24a.html | ICML 2024 | We propose MusicRL, the first music generation system finetuned from human feedback. Appreciation of text-to-music models is particularly subjective since the concept of musicality as well as the specific intention behind a caption are user-dependent (e.g. a caption such as “upbeat workout music” can map to a retro guitar solo or a technopop beat). Not only this makes supervised training of such models challenging, but it also calls for integrating continuous human feedback in their post-deployment finetuning. MusicRL is a pretrained autoregressive MusicLM model of discrete audio tokens finetuned with reinforcement learning to maximize sequence-level rewards. We design reward functions related specifically to text-adherence and audio quality with the help from selected raters, and use those to finetune MusicLM into MusicRL-R. We deploy MusicLM to users and collect a substantial dataset comprising 300,000 pairwise preferences. Using Reinforcement Learning from Human Feedback (RLHF), we train MusicRL-U, the first text-to-music model that incorporates human feedback at scale. Human evaluations show that both MusicRL-R and MusicRL-U are preferred to the baseline. Ultimately, MusicRL-RU combines the two approaches and results in the best model according to human raters. Ablation studies shed light on the musical attributes influencing human preferences, indicating that text adherence and quality only account for a part of it. This underscores the prevalence of subjectivity in musical appreciation and calls for further involvement of human listeners in the finetuning of music generation models. Samples can be found at google-research.github.io/seanet/musiclm/rlhf/. |
https://proceedings.mlr.press/v235/cini24a.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/cini24a/cini24a.pdf | https://openreview.net/forum?id=nd47Za5jk5 | Graph-based Time Series Clustering for End-to-End Hierarchical Forecasting | https://proceedings.mlr.press/v235/cini24a.html | Andrea Cini, Danilo Mandic, Cesare Alippi | https://proceedings.mlr.press/v235/cini24a.html | ICML 2024 | Relationships among time series can be exploited as inductive biases in learning effective forecasting models. In hierarchical time series, relationships among subsets of sequences induce hard constraints (hierarchical inductive biases) on the predicted values. In this paper, we propose a graph-based methodology to unify relational and hierarchical inductive biases in the context of deep learning for time series forecasting. In particular, we model both types of relationships as dependencies in a pyramidal graph structure, with each pyramidal layer corresponding to a level of the hierarchy. By exploiting modern - trainable - graph pooling operators we show that the hierarchical structure, if not available as a prior, can be learned directly from data, thus obtaining cluster assignments aligned with the forecasting objective. A differentiable reconciliation stage is incorporated into the processing architecture, allowing hierarchical constraints to act both as an architectural bias as well as a regularization element for predictions. Simulation results on representative datasets show that the proposed method compares favorably against the state of the art. |
https://proceedings.mlr.press/v235/clarke24a.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/clarke24a/clarke24a.pdf | https://openreview.net/forum?id=mK6FB9xQ7v | Studying K-FAC Heuristics by Viewing Adam through a Second-Order Lens | https://proceedings.mlr.press/v235/clarke24a.html | Ross M Clarke, José Miguel Hernández-Lobato | https://proceedings.mlr.press/v235/clarke24a.html | ICML 2024 | Research into optimisation for deep learning is characterised by a tension between the computational efficiency of first-order, gradient-based methods (such as SGD and Adam) and the theoretical efficiency of second-order, curvature-based methods (such as quasi-Newton methods and K-FAC). Noting that second-order methods often only function effectively with the addition of stabilising heuristics (such as Levenberg-Marquardt damping), we ask how much these (as opposed to the second-order curvature model) contribute to second-order algorithms’ performance. We thus study AdamQLR: an optimiser combining damping and learning rate selection techniques from K-FAC (Martens & Grosse, 2015) with the update directions proposed by Adam, inspired by considering Adam through a second-order lens. We evaluate AdamQLR on a range of regression and classification tasks at various scales and hyperparameter tuning methodologies, concluding K-FAC’s adaptive heuristics are of variable standalone general effectiveness, and finding an untuned AdamQLR setting can achieve comparable performance vs runtime to tuned benchmarks. |
https://proceedings.mlr.press/v235/clavier24a.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/clavier24a/clavier24a.pdf | https://openreview.net/forum?id=a1GvTbadqA | $\mathttVITS$ : Variational Inference Thompson Sampling for contextual bandits | https://proceedings.mlr.press/v235/clavier24a.html | Pierre Clavier, Tom Huix, Alain Oliviero Durmus | https://proceedings.mlr.press/v235/clavier24a.html | ICML 2024 | In this paper, we introduce and analyze a variant of the Thompson sampling (TS) algorithm for contextual bandits. At each round, traditional TS requires samples from the current posterior distribution, which is usually intractable. To circumvent this issue, approximate inference techniques can be used and provide samples with distribution close to the posteriors. However, current approximate techniques yield to either poor estimation (Laplace approximation) or can be computationally expensive (MCMC methods, Ensemble sampling...). In this paper, we propose a new algorithm, Varational Inference TS $\mathtt{VITS}$, based on Gaussian Variational Inference. This scheme provides powerful posterior approximations which are easy to sample from, and is computationally efficient, making it an ideal choice for TS. In addition, we show that $\mathtt{VITS}$ achieves a sub-linear regret bound of the same order in the dimension and number of round as traditional TS for linear contextual bandit. Finally, we demonstrate experimentally the effectiveness of $\mathtt{VITS}$ on both synthetic and real world datasets |
https://proceedings.mlr.press/v235/coda-forno24a.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/coda-forno24a/coda-forno24a.pdf | https://openreview.net/forum?id=Q3104y8djk | CogBench: a large language model walks into a psychology lab | https://proceedings.mlr.press/v235/coda-forno24a.html | Julian Coda-Forno, Marcel Binz, Jane X Wang, Eric Schulz | https://proceedings.mlr.press/v235/coda-forno24a.html | ICML 2024 | Large language models (LLMs) have significantly advanced the field of artificial intelligence. Yet, evaluating them comprehensively remains challenging. We argue that this is partly due to the predominant focus on performance metrics in most benchmarks. This paper introduces CogBench, a benchmark that includes ten behavioral metrics derived from seven cognitive psychology experiments. This novel approach offers a toolkit for phenotyping LLMs’ behavior. We apply CogBench to 40 LLMs, yielding a rich and diverse dataset. We analyze this data using statistical multilevel modeling techniques, accounting for the nested dependencies among fine-tuned versions of specific LLMs. Our study highlights the crucial role of model size and reinforcement learning from human feedback (RLHF) in improving performance and aligning with human behavior. Interestingly, we find that open-source models are less risk-prone than proprietary models and that fine-tuning on code does not necessarily enhance LLMs’ behavior. Finally, we explore the effects of prompt-engineering techniques. We discover that chain-of-thought prompting improves probabilistic reasoning, while take-a-step-back prompting fosters model-based behaviors. |
https://proceedings.mlr.press/v235/cohen24a.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/cohen24a/cohen24a.pdf | https://openreview.net/forum?id=kUm9iuvwIQ | Slicedit: Zero-Shot Video Editing With Text-to-Image Diffusion Models Using Spatio-Temporal Slices | https://proceedings.mlr.press/v235/cohen24a.html | Nathaniel Cohen, Vladimir Kulikov, Matan Kleiner, Inbar Huberman-Spiegelglas, Tomer Michaeli | https://proceedings.mlr.press/v235/cohen24a.html | ICML 2024 | Text-to-image (T2I) diffusion models achieve state-of-the-art results in image synthesis and editing. However, leveraging such pre-trained models for video editing is considered a major challenge. Many existing works attempt to enforce temporal consistency in the edited video through explicit correspondence mechanisms, either in pixel space or between deep features. These methods, however, struggle with strong nonrigid motion. In this paper, we introduce a fundamentally different approach, which is based on the observation that spatiotemporal slices of natural videos exhibit similar characteristics to natural images. Thus, the same T2I diffusion model that is normally used only as a prior on video frames, can also serve as a strong prior for enhancing temporal consistency by applying it on spatiotemporal slices. Based on this observation, we present Slicedit, a method for text-based video editing that utilizes a pre-trained T2I diffusion model to process both spatial and spatiotemporal slices. Our method generates videos that retain the structure and motion of the original video while adhering to the target text. Through extensive experiments, we demonstrate Slicedit’s ability to edit a wide range of real-world videos, confirming its clear advantages compared to existing baselines. |
https://proceedings.mlr.press/v235/cohen24b.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/cohen24b/cohen24b.pdf | https://openreview.net/forum?id=Lfp5Dk1xb6 | Improving Token-Based World Models with Parallel Observation Prediction | https://proceedings.mlr.press/v235/cohen24b.html | Lior Cohen, Kaixin Wang, Bingyi Kang, Shie Mannor | https://proceedings.mlr.press/v235/cohen24b.html | ICML 2024 | Motivated by the success of Transformers when applied to sequences of discrete symbols, token-based world models (TBWMs) were recently proposed as sample-efficient methods. In TBWMs, the world model consumes agent experience as a language-like sequence of tokens, where each observation constitutes a sub-sequence. However, during imagination, the sequential token-by-token generation of next observations results in a severe bottleneck, leading to long training times, poor GPU utilization, and limited representations. To resolve this bottleneck, we devise a novel Parallel Observation Prediction (POP) mechanism. POP augments a Retentive Network (RetNet) with a novel forward mode tailored to our reinforcement learning setting. We incorporate POP in a novel TBWM agent named REM (Retentive Environment Model), showcasing a 15.4x faster imagination compared to prior TBWMs. REM attains superhuman performance on 12 out of 26 games of the Atari 100K benchmark, while training in less than 12 hours. Our code is available at https://github.com/leor-c/REM |
https://proceedings.mlr.press/v235/cohen-addad24a.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/cohen-addad24a/cohen-addad24a.pdf | https://openreview.net/forum?id=45HNimd4YI | Perturb-and-Project: Differentially Private Similarities and Marginals | https://proceedings.mlr.press/v235/cohen-addad24a.html | Vincent Cohen-Addad, Tommaso D’Orsi, Alessandro Epasto, Vahab Mirrokni, Peilin Zhong | https://proceedings.mlr.press/v235/cohen-addad24a.html | ICML 2024 | We revisit the objective perturbations framework for differential privacy where noise is added to the input $A\in \mathcal{S}$ and the result is then projected back to the space of admissible datasets $\mathcal{S}$. Through this framework, we first design novel efficient algorithms to privately release pair-wise cosine similarities. Second, we derive a novel algorithm to compute $k$-way marginal queries over $n$ features. Prior work could achieve comparable guarantees only for $k$ even. Furthermore, we extend our results to $t$-sparse datasets, where our efficient algorithms yields novel, stronger guarantees whenever $t\le n^{5/6}/\log n.$ Finally, we provide a theoretical perspective on why fast input perturbation algorithms works well in practice. The key technical ingredients behind our results are tight sum-of-squares certificates upper bounding the Gaussian complexity of sets of solutions. |
https://proceedings.mlr.press/v235/cohen-addad24b.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/cohen-addad24b/cohen-addad24b.pdf | https://openreview.net/forum?id=BJx1K4lAAX | Multi-View Stochastic Block Models | https://proceedings.mlr.press/v235/cohen-addad24b.html | Vincent Cohen-Addad, Tommaso D’Orsi, Silvio Lattanzi, Rajai Nasser | https://proceedings.mlr.press/v235/cohen-addad24b.html | ICML 2024 | Graph clustering is a central topic in unsupervised learning with a multitude of practical applications. In recent years, multi-view graph clustering has gained a lot of attention for its applicability to real-world instances where one often has access to multiple data sources. In this paper we formalize a new family of models, called multi-view stochastic block models that capture this setting. For this model, we first study efficient algorithms that naively work on the union of multiple graphs. Then, we introduce a new efficient algorithm that provably outperforms previous approaches by analyzing the structure of each graph separately. Finally, we complement our results with an information-theoretic lower bound studying the limits of what can be done in this model. |
https://proceedings.mlr.press/v235/cohen-addad24c.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/cohen-addad24c/cohen-addad24c.pdf | https://openreview.net/forum?id=MSFxOMM0gK | A Near-Linear Time Approximation Algorithm for Beyond-Worst-Case Graph Clustering | https://proceedings.mlr.press/v235/cohen-addad24c.html | Vincent Cohen-Addad, Tommaso D’Orsi, Aida Mousavifar | https://proceedings.mlr.press/v235/cohen-addad24c.html | ICML 2024 | We consider the semi-random graph model of [Makarychev, Makarychev and Vijayaraghavan, STOC’12], where, given a random bipartite graph with $\alpha$ edges and an unknown bipartition $(A, B)$ of the vertex set, an adversary can add arbitrary edges inside each community and remove arbitrary edges from the cut $(A, B)$ (i.e. all adversarial changes are monotone with respect to the bipartition). For this model, a polynomial time algorithm [MMV’12] is known to approximate the Balanced Cut problem up to value $O(\alpha)$ as long as the cut $(A, B)$ has size $\Omega(\alpha)$. However, it consists of slow subroutines requiring optimal solutions for logarithmically many semidefinite programs. We study the fine-grained complexity of the problem and present the first near-linear time algorithm that achieves similar performances to that of [MMV’12]. Our algorithm runs in time $O(|V(G)|^{1+o(1)} + |E(G)|^{1+o(1)})$ and finds a balanced cut of value $O(\alpha).$ Our approach appears easily extendible to related problem, such as Sparsest Cut, and also yields an near-linear time $O(1)$-approximation to Dagupta’s objective function for hierarchical clustering [Dasgupta, STOC’16] for the semi-random hierarchical stochastic block model inputs of [Cohen-Addad, Kanade, Mallmann-Trenn, Mathieu, JACM’19]. |
https://proceedings.mlr.press/v235/cohen-addad24d.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/cohen-addad24d/cohen-addad24d.pdf | https://openreview.net/forum?id=3YG55Lbcnr | Dynamic Correlation Clustering in Sublinear Update Time | https://proceedings.mlr.press/v235/cohen-addad24d.html | Vincent Cohen-Addad, Silvio Lattanzi, Andreas Maggiori, Nikos Parotsidis | https://proceedings.mlr.press/v235/cohen-addad24d.html | ICML 2024 | We study the classic problem of correlation clustering in dynamic vertex streams. In this setting, vertices are either added or randomly deleted over time, and each vertex pair is connected by a positive or negative edge. The objective is to continuously find a partition which minimizes the sum of positive edges crossing clusters and negative edges within clusters. We present an algorithm that maintains an $O(1)$-approximation with $O(\text{polylog} n)$ amortized update time. Prior to our work Behnezhad et al. in SODA 2023 achieved a $5$-approximation with $O(1)$ expected update time in edge streams which translates in vertex streams to an $O(D)$-update time where $D$ is the maximum possible degree. Finally we complement our theoretical analysis with experiments on real world data. |
https://proceedings.mlr.press/v235/colbert24a.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/colbert24a/colbert24a.pdf | https://openreview.net/forum?id=mbx2pLK5Eq | A2Q+: Improving Accumulator-Aware Weight Quantization | https://proceedings.mlr.press/v235/colbert24a.html | Ian Colbert, Alessandro Pappalardo, Jakoba Petri-Koenig, Yaman Umuroglu | https://proceedings.mlr.press/v235/colbert24a.html | ICML 2024 | Quantization techniques commonly reduce the inference costs of neural networks by restricting the precision of weights and activations. Recent studies show that also reducing the precision of the accumulator can further improve hardware efficiency at the risk of numerical overflow, which introduces arithmetic errors that can degrade model accuracy. To avoid numerical overflow while maintaining accuracy, recent work proposed accumulator-aware quantization (A2Q)—a quantization-aware training method that constrains model weights during training to safely use a target accumulator bit width during inference. Although this shows promise, we demonstrate that A2Q relies on an overly restrictive constraint and a sub-optimal weight initialization strategy that each introduce superfluous quantization error. To address these shortcomings, we introduce: (1) an improved bound that alleviates accumulator constraints without compromising overflow avoidance; and (2) a new strategy for initializing quantized weights from pre-trained floating-point checkpoints. We combine these contributions with weight normalization to introduce A2Q+. We identify and characterize the various tradeoffs that arise as a consequence of accumulator constraints and support our analysis with experiments that show A2Q+ significantly improves these trade-offs when compared to prior methods. |
https://proceedings.mlr.press/v235/collins24a.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/collins24a/collins24a.pdf | https://openreview.net/forum?id=M8UbECx485 | Provable Multi-Task Representation Learning by Two-Layer ReLU Neural Networks | https://proceedings.mlr.press/v235/collins24a.html | Liam Collins, Hamed Hassani, Mahdi Soltanolkotabi, Aryan Mokhtari, Sanjay Shakkottai | https://proceedings.mlr.press/v235/collins24a.html | ICML 2024 | An increasingly popular machine learning paradigm is to pretrain a neural network (NN) on many tasks offline, then adapt it to downstream tasks, often by re-training only the last linear layer of the network. This approach yields strong downstream performance in a variety of contexts, demonstrating that multitask pretraining leads to effective feature learning. Although several recent theoretical studies have shown that shallow NNs learn meaningful features when either (i) they are trained on a single task or (ii) they are linear, very little is known about the closer-to-practice case of nonlinear NNs trained on multiple tasks. In this work, we present the first results proving that feature learning occurs during training with a nonlinear model on multiple tasks. Our key insight is that multi-task pretraining induces a pseudo-contrastive loss that favors representations that align points that typically have the same label across tasks. Using this observation, we show that when the tasks are binary classification tasks with labels depending on the projection of the data onto an $r$-dimensional subspace within the $d\gg r$-dimensional input space, a simple gradient-based multitask learning algorithm on a two-layer ReLU NN recovers this projection, allowing for generalization to downstream tasks with sample and neuron complexity independent of $d$. In contrast, we show that with high probability over the draw of a single task, training on this single task cannot guarantee to learn all $r$ ground-truth features. |
https://proceedings.mlr.press/v235/conitzer24a.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/conitzer24a/conitzer24a.pdf | https://openreview.net/forum?id=w1d9DOGymR | Position: Social Choice Should Guide AI Alignment in Dealing with Diverse Human Feedback | https://proceedings.mlr.press/v235/conitzer24a.html | Vincent Conitzer, Rachel Freedman, Jobst Heitzig, Wesley H. Holliday, Bob M. Jacobs, Nathan Lambert, Milan Mosse, Eric Pacuit, Stuart Russell, Hailey Schoelkopf, Emanuel Tewolde, William S. Zwicker | https://proceedings.mlr.press/v235/conitzer24a.html | ICML 2024 | Foundation models such as GPT-4 are fine-tuned to avoid unsafe or otherwise problematic behavior, such as helping to commit crimes or producing racist text. One approach to fine-tuning, called reinforcement learning from human feedback, learns from humans’ expressed preferences over multiple outputs. Another approach is constitutional AI, in which the input from humans is a list of high-level principles. But how do we deal with potentially diverging input from humans? How can we aggregate the input into consistent data about “collective” preferences or otherwise use it to make collective choices about model behavior? In this paper, we argue that the field of social choice is well positioned to address these questions, and we discuss ways forward for this agenda, drawing on discussions in a recent workshop on Social Choice for AI Ethics and Safety held in Berkeley, CA, USA in December 2023. |
https://proceedings.mlr.press/v235/cortes-gomez24a.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/cortes-gomez24a/cortes-gomez24a.pdf | https://openreview.net/forum?id=CiZN2OATRp | Statistical Inference Under Constrained Selection Bias | https://proceedings.mlr.press/v235/cortes-gomez24a.html | Santiago Cortes-Gomez, Mateo Dulce Rubio, Carlos Miguel Patiño, Bryan Wilder | https://proceedings.mlr.press/v235/cortes-gomez24a.html | ICML 2024 | Large-scale datasets are increasingly being used to inform decision making. While this effort aims to ground policy in real-world evidence, challenges have arisen as selection bias and other forms of distribution shifts often plague observational data. Previous attempts to provide robust inference have given guarantees depending on a user-specified amount of possible distribution shift (e.g., the maximum KL divergence between the observed and target distributions). However, decision makers will often have additional knowledge about the target distribution which constrains the kind of possible shifts. To leverage such information, we propose a framework that enables statistical inference in the presence of selection bias which obeys user-specified constraints in the form of functions whose expectation is known under the target distribution. The output is high-probability bounds on the value of an estimand for the target distribution. Hence, our method leverages domain knowledge in order to partially identify a wide class of estimands. We analyze the computational and statistical properties of methods to estimate these bounds and show that our method can produce informative bounds on a variety of simulated and semisynthetic tasks, as well as in a real-world use case. |
https://proceedings.mlr.press/v235/covert24a.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/covert24a/covert24a.pdf | https://openreview.net/forum?id=scSB9RynSd | Scaling Laws for the Value of Individual Data Points in Machine Learning | https://proceedings.mlr.press/v235/covert24a.html | Ian Connick Covert, Wenlong Ji, Tatsunori Hashimoto, James Zou | https://proceedings.mlr.press/v235/covert24a.html | ICML 2024 | Recent works have shown that machine learning models improve at a predictable rate with the amount of training data, leading to scaling laws that describe the relationship between error and dataset size. These scaling laws can help determine a model’s training dataset, but they take an aggregate view of the data by only considering the dataset’s size. We consider a new perspective by investigating scaling behavior for the value of individual data points: we find that a data point’s contribution to model’s performance shrinks predictably with the size of the dataset in a log-linear manner. Interestingly, there is significant variability in the scaling exponent among different data points, indicating that certain points are more valuable in small datasets and other points are relatively more useful as a part of large datasets. We provide learning theory support for our scaling laws and we observe empirically that it holds across several model classes. We further propose a maximum likelihood estimator and an amortized estimator to efficiently learn the individualized scaling behaviors from a small number of noisy observations per data point. Using our efficient estimators, we provide insights into factors that influence the scaling behavior of different data points. Finally we demonstrate applications of the individualized scaling laws to data valuation and data subset selection. |
https://proceedings.mlr.press/v235/crabbe24a.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/crabbe24a/crabbe24a.pdf | https://openreview.net/forum?id=W9GaJUVLCT | Time Series Diffusion in the Frequency Domain | https://proceedings.mlr.press/v235/crabbe24a.html | Jonathan Crabbé, Nicolas Huynh, Jan Pawel Stanczuk, Mihaela Van Der Schaar | https://proceedings.mlr.press/v235/crabbe24a.html | ICML 2024 | Fourier analysis has been an instrumental tool in the development of signal processing. This leads us to wonder whether this framework could similarly benefit generative modelling. In this paper, we explore this question through the scope of time series diffusion models. More specifically, we analyze whether representing time series in the frequency domain is a useful inductive bias for score-based diffusion models. By starting from the canonical SDE formulation of diffusion in the time domain, we show that a dual diffusion process occurs in the frequency domain with an important nuance: Brownian motions are replaced by what we call mirrored Brownian motions, characterized by mirror symmetries among their components. Building on this insight, we show how to adapt the denoising score matching approach to implement diffusion models in the frequency domain. This results in frequency diffusion models, which we compare to canonical time diffusion models. Our empirical evaluation on real-world datasets, covering various domains like healthcare and finance, shows that frequency diffusion models better capture the training distribution than time diffusion models. We explain this observation by showing that time series from these datasets tend to be more localized in the frequency domain than in the time domain, which makes them easier to model in the former case. All our observations point towards impactful synergies between Fourier analysis and diffusion models. |
https://proceedings.mlr.press/v235/cresswell24a.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/cresswell24a/cresswell24a.pdf | https://openreview.net/forum?id=4CO45y7Mlv | Conformal Prediction Sets Improve Human Decision Making | https://proceedings.mlr.press/v235/cresswell24a.html | Jesse C. Cresswell, Yi Sui, Bhargava Kumar, Noël Vouitsis | https://proceedings.mlr.press/v235/cresswell24a.html | ICML 2024 | In response to everyday queries, humans explicitly signal uncertainty and offer alternative answers when they are unsure. Machine learning models that output calibrated prediction sets through conformal prediction mimic this human behaviour; larger sets signal greater uncertainty while providing alternatives. In this work, we study the usefulness of conformal prediction sets as an aid for human decision making by conducting a pre-registered randomized controlled trial with conformal prediction sets provided to human subjects. With statistical significance, we find that when humans are given conformal prediction sets their accuracy on tasks improves compared to fixed-size prediction sets with the same coverage guarantee. The results show that quantifying model uncertainty with conformal prediction is helpful for human-in-the-loop decision making and human-AI teams. |
https://proceedings.mlr.press/v235/crispino24a.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/crispino24a/crispino24a.pdf | https://openreview.net/forum?id=zMwFvxr6CV | Agent Instructs Large Language Models to be General Zero-Shot Reasoners | https://proceedings.mlr.press/v235/crispino24a.html | Nicholas Crispino, Kyle Montgomery, Fankun Zeng, Dawn Song, Chenguang Wang | https://proceedings.mlr.press/v235/crispino24a.html | ICML 2024 | We introduce a method to improve the zero-shot reasoning abilities of large language models on general language understanding tasks. Specifically, we build an autonomous agent to instruct the reasoning process of large language models. To enable this, our agent only needs to generate a single set of instructions for each task. These instructions turn out to be extremely effective for improving the reasoning process of different large language models across all task instances. We show this approach further unleashes the zero-shot reasoning abilities of large language models to more tasks. We study the performance of our method on a wide set of datasets spanning generation, classification, and reasoning. We show that our method generalizes to most tasks and obtains state-of-the-art zero-shot performance on 20 of the 29 datasets that we evaluate. For instance, our method boosts the performance of state-of-the-art large language models by a large margin, including Vicuna-13b, Llama-2-70b-chat, and GPT-3.5 Turbo. Compared to zero-shot chain of thought, our improvement in reasoning is striking. With our method, Llama-2-70b-chat outperforms zero-shot GPT-3.5 Turbo significantly. |
https://proceedings.mlr.press/v235/crowson24a.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/crowson24a/crowson24a.pdf | https://openreview.net/forum?id=WRIn2HmtBS | Scalable High-Resolution Pixel-Space Image Synthesis with Hourglass Diffusion Transformers | https://proceedings.mlr.press/v235/crowson24a.html | Katherine Crowson, Stefan Andreas Baumann, Alex Birch, Tanishq Mathew Abraham, Daniel Z Kaplan, Enrico Shippole | https://proceedings.mlr.press/v235/crowson24a.html | ICML 2024 | We present the Hourglass Diffusion Transformer (HDiT), an image-generative model that exhibits linear scaling with pixel count, supporting training at high resolution (e.g. $1024 \times 1024$) directly in pixel-space. Building on the Transformer architecture, which is known to scale to billions of parameters, it bridges the gap between the efficiency of convolutional U-Nets and the scalability of Transformers. HDiT trains successfully without typical high-resolution training techniques such as multiscale architectures, latent autoencoders or self-conditioning. We demonstrate that HDiT performs competitively with existing models on ImageNet $256^2$, and sets a new state-of-the-art for diffusion models on FFHQ-$1024^2$. Code is available at https://github.com/crowsonkb/k-diffusion. |
https://proceedings.mlr.press/v235/csillag24a.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/csillag24a/csillag24a.pdf | https://openreview.net/forum?id=TejqrQBvll | Generalization Bounds for Causal Regression: Insights, Guarantees and Sensitivity Analysis | https://proceedings.mlr.press/v235/csillag24a.html | Daniel Csillag, Claudio Jose Struchiner, Guilherme Tegoni Goedert | https://proceedings.mlr.press/v235/csillag24a.html | ICML 2024 | Many algorithms have been recently proposed for causal machine learning. Yet, there is little to no theory on their quality, especially considering finite samples. In this work, we propose a theory based on generalization bounds that provides such guarantees. By introducing a novel change-of-measure inequality, we are able to tightly bound the model loss in terms of the deviation of the treatment propensities over the population, which we show can be empirically limited. Our theory is fully rigorous and holds even in the face of hidden confounding and violations of positivity. We demonstrate our bounds on semi-synthetic and real data, showcasing their remarkable tightness and practical utility. |
https://proceedings.mlr.press/v235/cui24a.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/cui24a/cui24a.pdf | https://openreview.net/forum?id=mslTE1qgLa | Major-Minor Mean Field Multi-Agent Reinforcement Learning | https://proceedings.mlr.press/v235/cui24a.html | Kai Cui, Christian Fabian, Anam Tahir, Heinz Koeppl | https://proceedings.mlr.press/v235/cui24a.html | ICML 2024 | Multi-agent reinforcement learning (MARL) remains difficult to scale to many agents. Recent MARL using Mean Field Control (MFC) provides a tractable and rigorous approach to otherwise difficult cooperative MARL. However, the strict MFC assumption of many independent, weakly-interacting agents is too inflexible in practice. We generalize MFC to instead simultaneously model many similar and few complex agents – as Major-Minor Mean Field Control (M3FC). Theoretically, we give approximation results for finite agent control, and verify the sufficiency of stationary policies for optimality together with a dynamic programming principle. Algorithmically, we propose Major-Minor Mean Field MARL (M3FMARL) for finite agent systems instead of the limiting system. The algorithm is shown to approximate the policy gradient of the underlying M3FC MDP. Finally, we demonstrate its capabilities experimentally in various scenarios. We observe a strong performance in comparison to state-of-the-art policy gradient MARL methods. |
https://proceedings.mlr.press/v235/cui24b.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/cui24b/cui24b.pdf | https://openreview.net/forum?id=o9uOuIwhZK | Learning Latent Space Hierarchical EBM Diffusion Models | https://proceedings.mlr.press/v235/cui24b.html | Jiali Cui, Tian Han | https://proceedings.mlr.press/v235/cui24b.html | ICML 2024 | This work studies the learning problem of the energy-based prior model and the multi-layer generator model. The multi-layer generator model, which contains multiple layers of latent variables organized in a top-down hierarchical structure, typically assumes the Gaussian prior model. Such a prior model can be limited in modelling expressivity, which results in a gap between the generator posterior and the prior model, known as the prior hole problem. Recent works have explored learning the energy-based (EBM) prior model as a second-stage, complementary model to bridge the gap. However, the EBM defined on a multi-layer latent space can be highly multi-modal, which makes sampling from such marginal EBM prior challenging in practice, resulting in ineffectively learned EBM. To tackle the challenge, we propose to leverage the diffusion probabilistic scheme to mitigate the burden of EBM sampling and thus facilitate EBM learning. Our extensive experiments demonstrate a superior performance of our diffusion-learned EBM prior on various challenging tasks. |
https://proceedings.mlr.press/v235/cui24c.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/cui24c/cui24c.pdf | https://openreview.net/forum?id=YYwERRXsJW | Harmonizing Generalization and Personalization in Federated Prompt Learning | https://proceedings.mlr.press/v235/cui24c.html | Tianyu Cui, Hongxia Li, Jingya Wang, Ye Shi | https://proceedings.mlr.press/v235/cui24c.html | ICML 2024 | Federated Prompt Learning (FPL) incorporates large pre-trained Vision-Language models (VLM) into federated learning through prompt tuning. The transferable representations and remarkable generalization capacity of VLM make them highly compatible with the integration of federated learning. Addressing data heterogeneity in federated learning requires personalization, but excessive focus on it across clients could compromise the model’s ability to generalize effectively. To preserve the impressive generalization capability of VLM, it is crucial to strike a balance between personalization and generalization in FPL. To tackle this challenge, we proposed Federated Prompt Learning with CLIP Generalization and low-rank Personalization (FedPGP), which employs pre-trained CLIP to provide knowledge-guidance on the global prompt for improved generalization and incorporates a low-rank adaptation term to personalize the global prompt. Further, FedPGP integrates a prompt-wise contrastive loss to achieve knowledge guidance and personalized adaptation simultaneously, enabling a harmonious balance between personalization and generalization in FPL. We conduct extensive experiments on various datasets to explore base-to-novel generalization in both category-level and domain-level scenarios with heterogeneous data, showing the superiority of FedPGP in balancing generalization and personalization. |
https://proceedings.mlr.press/v235/cui24d.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/cui24d/cui24d.pdf | https://openreview.net/forum?id=EdRb84fiJY | Asymptotics of feature learning in two-layer networks after one gradient-step | https://proceedings.mlr.press/v235/cui24d.html | Hugo Cui, Luca Pesce, Yatin Dandi, Florent Krzakala, Yue Lu, Lenka Zdeborova, Bruno Loureiro | https://proceedings.mlr.press/v235/cui24d.html | ICML 2024 | In this manuscript, we investigate the problem of how two-layer neural networks learn features from data, and improve over the kernel regime, after being trained with a single gradient descent step. Leveraging the insight from (Ba et al., 2022), we model the trained network by a spiked Random Features (sRF) model. Further building on recent progress on Gaussian universality (Dandi et al., 2023), we provide an exact asymptotic description of the generalization error of the sRF in the high-dimensional limit where the number of samples, the width, and the input dimension grow at a proportional rate. The resulting characterization for sRFs also captures closely the learning curves of the original network model. This enables us to understand how adapting to the data is crucial for the network to efficiently learn non-linear functions in the direction of the gradient - where at initialization it can only express linear functions in this regime. |
https://proceedings.mlr.press/v235/cui24e.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/cui24e/cui24e.pdf | https://openreview.net/forum?id=RbnojVv4HK | Ameliorate Spurious Correlations in Dataset Condensation | https://proceedings.mlr.press/v235/cui24e.html | Justin Cui, Ruochen Wang, Yuanhao Xiong, Cho-Jui Hsieh | https://proceedings.mlr.press/v235/cui24e.html | ICML 2024 | Dataset Condensation has emerged as a technique for compressing large datasets into smaller synthetic counterparts, facilitating downstream training tasks. In this paper, we study the impact of bias inside the original dataset on the performance of dataset condensation. With a comprehensive empirical evaluation on canonical datasets with color, corruption and background biases, we found that color and background biases in the original dataset will be amplified through the condensation process, resulting in a notable decline in the performance of models trained on the condensed dataset, while corruption bias is suppressed through the condensation process. To reduce bias amplification in dataset condensation, we introduce a simple yet highly effective approach based on a sample reweighting scheme utilizing kernel density estimation. Empirical results on multiple real-world and synthetic datasets demonstrate the effectiveness of the proposed method. Notably, on CMNIST with 5% bias-conflict ratio and IPC 50, our method achieves 91.5% test accuracy compared to 23.8% from vanilla DM, boosting the performance by 67.7%, whereas applying state-of-the-art debiasing method on the same dataset only achieves 53.7% accuracy. Our findings highlight the importance of addressing biases in dataset condensation and provide a promising avenue to address bias amplification in the process. |
https://proceedings.mlr.press/v235/cui24f.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/cui24f/cui24f.pdf | https://openreview.net/forum?id=BOorDpKHiJ | ULTRAFEEDBACK: Boosting Language Models with Scaled AI Feedback | https://proceedings.mlr.press/v235/cui24f.html | Ganqu Cui, Lifan Yuan, Ning Ding, Guanming Yao, Bingxiang He, Wei Zhu, Yuan Ni, Guotong Xie, Ruobing Xie, Yankai Lin, Zhiyuan Liu, Maosong Sun | https://proceedings.mlr.press/v235/cui24f.html | ICML 2024 | Learning from human feedback has become a pivot technique in aligning large language models (LLMs) with human preferences. However, acquiring vast and premium human feedback is bottlenecked by time, labor, and human capability, resulting in small sizes or limited topics of current datasets. This further hinders feedback learning as well as alignment research within the open-source community. To address this issue, we explore how to go beyond human feedback and collect high-quality AI feedback automatically for a scalable alternative. Specifically, we identify scale and diversity as the key factors for feedback data to take effect. Accordingly, we first broaden instructions and responses in both amount and breadth to encompass a wider range of user-assistant interactions. Then, we meticulously apply a series of techniques to mitigate annotation biases for more reliable AI feedback. We finally present UltraFeedback, a large-scale, high-quality, and diversified AI feedback dataset, which contains over 1 million GPT-4 feedback for 250k user-assistant conversations from various aspects. Built upon UltraFeedback, we align a LLaMA-based model by best-of-$n$ sampling and reinforcement learning, demonstrating its exceptional performance on chat benchmarks. Our work validates the effectiveness of scaled AI feedback data in constructing strong open-source chat language models, serving as a solid foundation for future feedback learning research. |
https://proceedings.mlr.press/v235/cullen24a.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/cullen24a/cullen24a.pdf | https://openreview.net/forum?id=RKlmOBFwAh | Et Tu Certifications: Robustness Certificates Yield Better Adversarial Examples | https://proceedings.mlr.press/v235/cullen24a.html | Andrew Craig Cullen, Shijie Liu, Paul Montague, Sarah Monazam Erfani, Benjamin I. P. Rubinstein | https://proceedings.mlr.press/v235/cullen24a.html | ICML 2024 | In guaranteeing the absence of adversarial examples in an instance’s neighbourhood, certification mechanisms play an important role in demonstrating neural net robustness. In this paper, we ask if these certifications can compromise the very models they help to protect? Our new Certification Aware Attack exploits certifications to produce computationally efficient norm-minimising adversarial examples $74$% more often than comparable attacks, while reducing the median perturbation norm by more than $10$%. While these attacks can be used to assess the tightness of certification bounds, they also highlight that releasing certifications can paradoxically reduce security. |
https://proceedings.mlr.press/v235/cyffers24a.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/cyffers24a/cyffers24a.pdf | https://openreview.net/forum?id=k2dVVIWWho | Differentially Private Decentralized Learning with Random Walks | https://proceedings.mlr.press/v235/cyffers24a.html | Edwige Cyffers, Aurélien Bellet, Jalaj Upadhyay | https://proceedings.mlr.press/v235/cyffers24a.html | ICML 2024 | The popularity of federated learning comes from the possibility of better scalability and the ability for participants to keep control of their data, improving data security and sovereignty. Unfortunately, sharing model updates also creates a new privacy attack surface. In this work, we characterize the privacy guarantees of decentralized learning with random walk algorithms, where a model is updated by traveling from one node to another along the edges of a communication graph. Using a recent variant of differential privacy tailored to the study of decentralized algorithms, namely Pairwise Network Differential Privacy, we derive closed-form expressions for the privacy loss between each pair of nodes where the impact of the communication topology is captured by graph theoretic quantities. Our results further reveal that random walk algorithms tends to yield better privacy guarantees than gossip algorithms for nodes close from each other. We supplement our theoretical results with empirical evaluation on synthetic and real-world graphs and datasets. |
https://proceedings.mlr.press/v235/dagan24a.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/dagan24a/dagan24a.pdf | https://openreview.net/forum?id=ZFYBnLljtT | Getting the most out of your tokenizer for pre-training and domain adaptation | https://proceedings.mlr.press/v235/dagan24a.html | Gautier Dagan, Gabriel Synnaeve, Baptiste Roziere | https://proceedings.mlr.press/v235/dagan24a.html | ICML 2024 | Tokenization is an understudied and often neglected component of modern LLMs. Most published works use a single tokenizer for all experiments, often borrowed from another model, without performing ablations or analysis to optimize tokenization. Moreover, the tokenizer is generally kept unchanged when fine-tuning a base model. In this paper, we show that the size, pre-tokenization regular expression, and training data of a tokenizer can significantly impact the model’s generation speed, effective context size, memory usage, and downstream performance. We train specialized Byte-Pair Encoding code tokenizers, and conduct extensive ablations on the impact of tokenizer design on the performance of LLMs for code generation tasks such as HumanEval and MBPP, and provide recommendations for tokenizer hyper-parameters selection and switching the tokenizer in a pre-trained LLM. We perform our experiments on models trained from scratch and from pre-trained models, verifying their applicability to a wide range of use-cases. We find that when fine-tuning on more than 50 billion tokens, we can specialize the tokenizer of a pre-trained LLM to obtain large gains in generation speed and effective context size. |
https://proceedings.mlr.press/v235/dahan24a.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/dahan24a/dahan24a.pdf | https://openreview.net/forum?id=Ht20wtgaty | Fault Tolerant ML: Efficient Meta-Aggregation and Synchronous Training | https://proceedings.mlr.press/v235/dahan24a.html | Tehila Dahan, Kfir Yehuda Levy | https://proceedings.mlr.press/v235/dahan24a.html | ICML 2024 | In this paper, we investigate the challenging framework of Byzantine-robust training in distributed machine learning (ML) systems, focusing on enhancing both efficiency and practicality. As distributed ML systems become integral for complex ML tasks, ensuring resilience against Byzantine failures—where workers may contribute incorrect updates due to malice or error—gains paramount importance. Our first contribution is the introduction of the Centered Trimmed Meta Aggregator (CTMA), an efficient meta-aggregator that upgrades baseline aggregators to optimal performance levels, while requiring low computational demands. Additionally, we propose harnessing a recently developed gradient estimation technique based on a double-momentum strategy within the Byzantine context. Our paper highlights its theoretical and practical advantages for Byzantine-robust training, especially in simplifying the tuning process and reducing the reliance on numerous hyperparameters. The effectiveness of this technique is supported by theoretical insights within the stochastic convex optimization (SCO) framework and corroborated by empirical evidence. |
https://proceedings.mlr.press/v235/dai24a.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/dai24a/dai24a.pdf | https://openreview.net/forum?id=4XlGXIh2BB | Position: Beyond Personhood: Agency, Accountability, and the Limits of Anthropomorphic Ethical Analysis | https://proceedings.mlr.press/v235/dai24a.html | Jessica Dai | https://proceedings.mlr.press/v235/dai24a.html | ICML 2024 | What is agency, and why does it matter? In this work, we draw from the political science and philosophy literature and give two competing visions of what it means to be an (ethical) agent. The first view, which we term mechanistic, is commonly— and implicitly—assumed in AI research, yet it is a fundamentally limited means to understand the ethical characteristics of AI. Under the second view, which we term volitional, AI can no longer be considered an ethical agent. We discuss the implications of each of these views for two critical questions: first, what the ideal system “ought” to look like, and second, how accountability may be achieved. In light of this discussion, we ultimately argue that, in the context of ethically-significant behavior, AI should be viewed not as an agent but as the outcome of political processes. |
https://proceedings.mlr.press/v235/dai24b.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/dai24b/dai24b.pdf | https://openreview.net/forum?id=uEx2bSAJu8 | Multi-View Clustering by Inter-cluster Connectivity Guided Reward | https://proceedings.mlr.press/v235/dai24b.html | Hao Dai, Yang Liu, Peng Su, Hecheng Cai, Shudong Huang, Jiancheng Lv | https://proceedings.mlr.press/v235/dai24b.html | ICML 2024 | Multi-view clustering has been widely explored for its effectiveness in harmonizing heterogeneity along with consistency in different views of data. Despite the significant progress made by recent works, the performance of most existing methods is heavily reliant on strong priori information regarding the true cluster number $\textit{K}$, which is rarely feasible in real-world scenarios. In this paper, we propose a novel graph-based multi-view clustering algorithm to infer unknown $\textit{K}$ through a graph consistency reward mechanism. To be specific, we evaluate the cluster indicator matrix during each iteration with respect to diverse $\textit{K}$. We formulate the inference process of unknown $\textit{K}$ as a parsimonious reinforcement learning paradigm, where the reward is measured by inter-cluster connectivity. As a result, our approach is capable of independently producing the final clustering result, free from the input of a predefined cluster number. Experimental results on multiple benchmark datasets demonstrate the effectiveness of our proposed approach in comparison to existing state-of-the-art methods. |
https://proceedings.mlr.press/v235/dai24c.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/dai24c/dai24c.pdf | https://openreview.net/forum?id=PDO2Oc1cS1 | High-Order Contrastive Learning with Fine-grained Comparative Levels for Sparse Ordinal Tensor Completion | https://proceedings.mlr.press/v235/dai24c.html | Yu Dai, Junchen Shen, Zijie Zhai, Danlin Liu, Jingyang Chen, Yu Sun, Ping Li, Jie Zhang, Kai Zhang | https://proceedings.mlr.press/v235/dai24c.html | ICML 2024 | Contrastive learning is a powerful paradigm for representation learning with prominent success in computer vision and NLP, but how to extend its success to high-dimensional tensors remains a challenge. This is because tensor data often exhibit high-order mode-interactions that are hard to profile and with negative samples growing combinatorially faster than second-order contrastive learning; furthermore, many real-world tensors have ordinal entries that necessitate more delicate comparative levels. To solve the challenge, we propose High-Order Contrastive Tensor Completion (HOCTC), an innovative network to extend contrastive learning to sparse ordinal tensor data. HOCTC employs a novel attention-based strategy with query-expansion to capture high-order mode interactions even in case of very limited tokens, which transcends beyond second-order learning scenarios. Besides, it extends two-level comparisons (positive-vs-negative) to fine-grained contrast-levels using ordinal tensor entries as a natural guidance. Efficient sampling scheme is proposed to enforce such delicate comparative structures, generating comprehensive self-supervised signals for high-order representation learning. Extensive experiments show that HOCTC has promising results in sparse tensor completion in traffic/recommender applications. |
https://proceedings.mlr.press/v235/dai24d.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/dai24d/dai24d.pdf | https://openreview.net/forum?id=BiENLaUwlK | Safe Reinforcement Learning using Finite-Horizon Gradient-based Estimation | https://proceedings.mlr.press/v235/dai24d.html | Juntao Dai, Yaodong Yang, Qian Zheng, Gang Pan | https://proceedings.mlr.press/v235/dai24d.html | ICML 2024 | A key aspect of Safe Reinforcement Learning (Safe RL) involves estimating the constraint condition for the next policy, which is crucial for guiding the optimization of safe policy updates. However, the existing Advantage-based Estimation (ABE) method relies on the infinite-horizon discounted advantage function. This dependence leads to catastrophic errors in finite-horizon scenarios with non-discounted constraints, resulting in safety-violation updates. In response, we propose the first estimation method for finite-horizon non-discounted constraints in deep Safe RL, termed Gradient-based Estimation (GBE), which relies on the analytic gradient derived along trajectories. Our theoretical and empirical analyses demonstrate that GBE can effectively estimate constraint changes over a finite horizon. Constructing a surrogate optimization problem with GBE, we developed a novel Safe RL algorithm called Constrained Gradient-based Policy Optimization (CGPO). CGPO identifies feasible optimal policies by iteratively resolving sub-problems within trust regions. Our empirical results reveal that CGPO, unlike baseline algorithms, successfully estimates the constraint functions of subsequent policies, thereby ensuring the efficiency and feasibility of each update. |
https://proceedings.mlr.press/v235/daley24a.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/daley24a/daley24a.pdf | https://openreview.net/forum?id=jM9A3Kz6Ki | Averaging $n$-step Returns Reduces Variance in Reinforcement Learning | https://proceedings.mlr.press/v235/daley24a.html | Brett Daley, Martha White, Marlos C. Machado | https://proceedings.mlr.press/v235/daley24a.html | ICML 2024 | Multistep returns, such as $n$-step returns and $\lambda$-returns, are commonly used to improve the sample efficiency of reinforcement learning (RL) methods. The variance of the multistep returns becomes the limiting factor in their length; looking too far into the future increases variance and reverses the benefits of multistep learning. In our work, we demonstrate the ability of compound returns—weighted averages of $n$-step returns—to reduce variance. We prove for the first time that any compound return with the same contraction modulus as a given $n$-step return has strictly lower variance. We additionally prove that this variance-reduction property improves the finite-sample complexity of temporal-difference learning under linear function approximation. Because general compound returns can be expensive to implement, we introduce two-bootstrap returns which reduce variance while remaining efficient, even when using minibatched experience replay. We conduct experiments showing that compound returns often increase the sample efficiency of $n$-step deep RL agents like DQN and PPO. |
https://proceedings.mlr.press/v235/dalirrooyfard24a.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/dalirrooyfard24a/dalirrooyfard24a.pdf | https://openreview.net/forum?id=saP7s0ZgYE | Pruned Pivot: Correlation Clustering Algorithm for Dynamic, Parallel, and Local Computation Models | https://proceedings.mlr.press/v235/dalirrooyfard24a.html | Mina Dalirrooyfard, Konstantin Makarychev, Slobodan Mitrovic | https://proceedings.mlr.press/v235/dalirrooyfard24a.html | ICML 2024 | Given a graph with positive and negative edge labels, the correlation clustering problem aims to cluster the nodes so to minimize the total number of between-cluster positive and within-cluster negative edges. This problem has many applications in data mining, particularly in unsupervised learning. Inspired by the prevalence of large graphs and constantly changing data in modern applications, we study correlation clustering in dynamic, parallel (MPC), and local computation (LCA) settings. We design an approach that improves state-of-the-art runtime complexities in all these settings. In particular, we provide the first fully dynamic algorithm that runs in an expected amortized constant time, without any dependence on the graph size. Moreover, our algorithm essentially matches the approximation guarantee of the celebrated Pivot algorithm. |
https://proceedings.mlr.press/v235/dalton24a.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/dalton24a/dalton24a.pdf | https://openreview.net/forum?id=1V50J0emll | Physics and Lie symmetry informed Gaussian processes | https://proceedings.mlr.press/v235/dalton24a.html | David Dalton, Dirk Husmeier, Hao Gao | https://proceedings.mlr.press/v235/dalton24a.html | ICML 2024 | Physics-informed machine learning (PIML) has established itself as a new scientific paradigm which enables the seamless integration of observational data with partial differential equation (PDE) based physics models. A powerful tool for the analysis, reduction and solution of PDEs is the Lie symmetry method. Nevertheless, only recently has the integration of such symmetries into PIML frameworks begun to be explored. The present work adds to this growing literature by introducing an approach for incorporating a Lie symmetry into a physics-informed Gaussian process (GP) model. The symmetry is introduced as a constraint on the GP; either in a soft manner via virtual observations of an induced PDE called the invariant surface condition, or explicitly through the design of the kernel. Experimental results demonstrate that the use of symmetry constraints improves the performance of the GP for both forward and inverse problems, and that our approach offers competitive performance with neural networks in the low-data environment. |
https://proceedings.mlr.press/v235/dan24a.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/dan24a/dan24a.pdf | https://openreview.net/forum?id=EYOo48YGhy | Exploring the Enigma of Neural Dynamics Through A Scattering-Transform Mixer Landscape for Riemannian Manifold | https://proceedings.mlr.press/v235/dan24a.html | Tingting Dan, Ziquan Wei, Won Hwa Kim, Guorong Wu | https://proceedings.mlr.press/v235/dan24a.html | ICML 2024 | The human brain is a complex inter-wired system that emerges spontaneous functional fluctuations. In spite of tremendous success in the experimental neuroscience field, a system-level understanding of how brain anatomy supports various neural activities remains elusive. Capitalizing on the unprecedented amount of neuroimaging data, we present a physics-informed deep model to uncover the coupling mechanism between brain structure and function through the lens of data geometry that is rooted in the widespread wiring topology of connections between distant brain regions. Since deciphering the puzzle of self-organized patterns in functional fluctuations is the gateway to understanding the emergence of cognition and behavior, we devise a geometric deep model to uncover manifold mapping functions that characterize the intrinsic feature representations of evolving functional fluctuations on the Riemannian manifold. In lieu of learning unconstrained mapping functions, we introduce a set of graph-harmonic scattering transforms to impose the brain-wide geometry on top of manifold mapping functions, which allows us to cast the manifold-based deep learning into a reminiscent of MLP-Mixer architecture (in computer vision) for Riemannian manifold. As a proof-of-concept approach, we explore a neural-manifold perspective to understand the relationship between (static) brain structure and (dynamic) function, challenging the prevailing notion in cognitive neuroscience by proposing that neural activities are essentially excited by brain-wide oscillation waves living on the geometry of human connectomes, instead of being confined to focal areas. |
https://proceedings.mlr.press/v235/dandi24a.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/dandi24a/dandi24a.pdf | https://openreview.net/forum?id=iKkFruh4d5 | The Benefits of Reusing Batches for Gradient Descent in Two-Layer Networks: Breaking the Curse of Information and Leap Exponents | https://proceedings.mlr.press/v235/dandi24a.html | Yatin Dandi, Emanuele Troiani, Luca Arnaboldi, Luca Pesce, Lenka Zdeborova, Florent Krzakala | https://proceedings.mlr.press/v235/dandi24a.html | ICML 2024 | We investigate the training dynamics of two-layer neural networks when learning multi-index target functions. We focus on multi-pass gradient descent (GD) that reuses the batches multiple times and show that it significantly changes the conclusion about which functions are learnable compared to single-pass gradient descent. In particular, multi-pass GD with finite stepsize is found to overcome the limitations of gradient flow and single-pass GD given by the information exponent (Ben Arous et al., 2021) and leap exponent (Abbe et al., 2023) of the target function. We show that upon re-using batches, the network achieves in just two time steps an overlap with the target subspace even for functions not satisfying the staircase property (Abbe et al., 2021). We characterize the (broad) class of functions efficiently learned in finite time. The proof of our results is based on the analysis of the Dynamical Mean-Field Theory (DMFT). We further provide a closed-form description of the dynamical process of the low-dimensional projections of the weights, and numerical experiments illustrating the theory. |
https://proceedings.mlr.press/v235/dang24a.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/dang24a/dang24a.pdf | https://openreview.net/forum?id=YBetKvUlF7 | Neural Collapse for Cross-entropy Class-Imbalanced Learning with Unconstrained ReLU Features Model | https://proceedings.mlr.press/v235/dang24a.html | Hien Dang, Tho Tran Huu, Tan Minh Nguyen, Nhat Ho | https://proceedings.mlr.press/v235/dang24a.html | ICML 2024 | The current paradigm of training deep neural networks for classification tasks includes minimizing the empirical risk, pushing the training loss value towards zero even after the training classification error has vanished. In this terminal phase of training, it has been observed that the last-layer features collapse to their class-means and these class-means converge to the vertices of a simplex Equiangular Tight Frame (ETF). This phenomenon is termed as Neural Collapse ($\mathcal{NC}$). However, this characterization only holds in class-balanced datasets where every class has the same number of training samples. When the training dataset is class-imbalanced, some $\mathcal{NC}$ properties will no longer hold true, for example, the geometry of class-means will skew away from the simplex ETF. In this paper, we generalize $\mathcal{NC}$ to imbalanced regime for cross-entropy loss under the unconstrained ReLU features model. We demonstrate that while the within-class features collapse property still holds in this setting, the class-means will converge to a structure consisting of orthogonal vectors with lengths dependent on the number of training samples. Furthermore, we find that the classifier weights (i.e., the last-layer linear classifier) are aligned to the scaled and centered class-means, with scaling factors dependent on the number of training samples of each class. This generalizes $\mathcal{NC}$ in the class-balanced setting. We empirically validate our results through experiments on practical architectures and dataset. |
https://proceedings.mlr.press/v235/dao24a.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/dao24a/dao24a.pdf | https://openreview.net/forum?id=ztn8FCR1td | Transformers are SSMs: Generalized Models and Efficient Algorithms Through Structured State Space Duality | https://proceedings.mlr.press/v235/dao24a.html | Tri Dao, Albert Gu | https://proceedings.mlr.press/v235/dao24a.html | ICML 2024 | While Transformers have been the main architecture behind deep learning’s success in language modeling, state-space models (SSMs) such as Mamba have recently been shown to match or outperform Transformers at small to medium scale. We show that these families of models are actually quite closely related, and develop a rich framework of theoretical connections between SSMs and variants of attention, connected through various decompositions of a well-studied class of structured semiseparable matrices. Our state space duality (SSD) framework allows us to design a new architecture (Mamba-2) whose core layer is an a refinement of Mamba’s selective SSM that is 2-8$\times$ faster, while continuing to be competitive with Transformers on language modeling. |
https://proceedings.mlr.press/v235/dao24b.html | https://raw.githubusercontent.com/mlresearch/v235/main/assets/dao24b/dao24b.pdf | https://openreview.net/forum?id=aLSA3JH08h | Boosting Offline Optimizers with Surrogate Sensitivity | https://proceedings.mlr.press/v235/dao24b.html | Manh Cuong Dao, Phi Le Nguyen, Thao Nguyen Truong, Trong Nghia Hoang | https://proceedings.mlr.press/v235/dao24b.html | ICML 2024 | Offline optimization is an important task in numerous material engineering domains where online experimentation to collect data is too expensive and needs to be replaced by an in silico maximization of a surrogate of the black-box function. Although such a surrogate can be learned from offline data, its prediction might not be reliable outside the offline data regime, which happens when the surrogate has narrow prediction margin and is (therefore) sensitive to small perturbations of its parameterization. This raises the following questions: (1) how to regulate the sensitivity of a surrogate model; and (2) whether conditioning an offline optimizer with such less sensitive surrogate will lead to better optimization performance. To address these questions, we develop an optimizable sensitivity measurement for the surrogate model, which then inspires a sensitivity-informed regularizer that is applicable to a wide range of offline optimizers. This development is both orthogonal and synergistic to prior research on offline optimization, which is demonstrated in our extensive experiment benchmark. |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.