CS59200: Probabilistic Machine Learning (Fall 2022)

How can a machine learn from data? How should we construct appropriate models? How do we design algorithms for inference? What are the fundamental design principles of data-driven learning and decision-making?

This course will help you answer these fundamental questions in machine learning from a probabilistic perspective. You will read and discuss research papers in probabilistic machine learning, including but not limited to research topics about

• • Probabilistic models: Bayesian neural networks, variational autoencoder, normalizing flows, score-based generative models, diffusion models, Gaussian processes.
• • Probabilistic inference: Markov chain Monte Carlo (MCMC), variational inference, Laplace approximation, dropout, ensemble.
• • Evaluations and applications: uncertainty estimation, calibration, distribution shift, out-of-distribution detection, adversarial attacks.