scottymcgee/hw2_tabular

Model Card for Mozart vs Beethoven Classifier

This model predicts whether a classical piano piece was composed by Mozart or Beethoven, based on numerical features extracted from the score (counts of right-hand notes, left-hand notes, measures, key centers, and markings).

Model Details

Model Description

• Developed by: Scotty McGee (PhD student, CMU)
• Shared by [optional]: Scotty McGee
• Model type: Tabular classification (AutoML with AutoGluon)
• Language(s) (NLP): Not applicable (tabular/numeric features only)
• License: MIT (update as needed)
• Finetuned from model: N/A

Uses

Direct Use

Demonstration of machine-learning classification on musical data. Predicts a binary composer label (Mozart or Beethoven) from numeric score features.

Downstream Use [optional]

Could be adapted for broader composer classification tasks, musicology studies, or automated metadata tagging.

Out-of-Scope Use

• Not intended as a general music recognition tool.
• Not reliable for real performance or music audio classification.
• Not suitable for commercial music rights enforcement.

Bias, Risks, and Limitations

• Limited to Mozart and Beethoven; not generalizable to other composers.
• Features are simplistic (counts of notes, measures, key centers, markings).
• May not capture stylistic nuance.
• Risk of overfitting to the dataset used.

Recommendations

Use for small-scale experiments and demos. Do not apply to large-scale music classification tasks without retraining and validation.

How to Get Started with the Model

from autogluon.tabular import TabularPredictor
preds = predictor.predict(df_test)  # Mozart or Beethoven