
Feature Extraction and Machine Learning (III): Artificial Intelligence
Michael Scott Asato Cuthbert continues his MIT course 21M.383, Computational Music Theory and Analysis, with the third session on feature extraction and machine learning applied to music. He covers how numeric features are pulled from musical scores and audio, how those features feed into machine learning models, and how the music21 toolkit implements this pipeline in practice. The lecture works through artificial intelligence concepts as they apply to musical data, connecting general machine learning ideas, classification, training data, model evaluation, to concrete coding examples in music21. Cuthbert demonstrates extracting features from score objects and discusses how computational analysis can support tasks like style classification or corpus study. The session assumes familiarity with earlier lectures in the series on feature extraction basics and builds toward a working understanding of how machine learning techniques can be used to analyze musical structure computationally.