CSE 190

Instructor: Shlomo Dubnov

Quarters: Summer 2024 and Summer 2025

This special topics course introduces students to the ways in which AI and ML are used in the field of music and audio. The course begins with the basics of signal processing, music theory, the discrete fourier transform, and spectrograms. Then, some foundational algorithms such as Griffin-Lim reconstruction and Mel-frequency cepstral coefficients (MFCCs) are discussed. The course then transitions to deep-learning techniques such as Autoencoder denoising, generating MIDI files using RNNs, genre classification using CNNs and RNNs, and using GANs to generate music. The course contains programming assignments that give hands-on experience with these algorithms and Python packages for audio processing such as librosa, pyaudio, pretty-midi, vmo, music21, jchord, etc. in addition to TensorFlow and Keras for deep learning.

I helped recreate many of the programming assignments for this course using the latest versions of Tensorflow to allow compatibility with platforms such as Google Colab.

The website contains all the course logistics and materials.