Learning and composing of classical music using restricted Boltzmann machines

The episode discusses a paper by Kobayashi and Watanabe on using restricted Boltzmann machines (RBMs) to learn and compose classical music from piano roll images of Bach and Mozart. The hosts explore how the simple model can generate two-measure coherent pieces but fails at longer structures. They conclude that while the model learns rhythm, its internal representations do not map onto human music theory.

Open the full episode

More from AI Radio