File size: 858 Bytes
b118f86 7377a42 c0e5193 b118f86 1240765 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 |
---
title: RL Interpretable Policy Via Kolmogorov Arnold Network
emoji: 🧠➡️🔢
colorFrom: red
colorTo: purple
sdk: gradio
sdk_version: 4.29.0
app_file: app.py
pinned: false
license: mit
---
Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
### Application demo :
- Choose a RL environment from the gymnasium library. A policy from a pre-trained Proximal Policy Optimization (PPO) agent will automatically be loaded, which generates an expert dataset and videos of the agent's performance in the selected environment.
- Click the "Compute Symbolic Policy" button to train a KAN policy on the expert dataset. Once it is done, you can visualize the KAN network and watch videos of the KAN agent's performance in the selected environment !
<img alt="Interpretability app demo" src="demo/app_demo.gif">
|