Hide Advanced Options
Courses - Fall 2024
PHYS
Physics Department Site
Open Seats as of
12/17/2024 at 10:30 PM
PHYS486
(Perm Req)
Machine Learning for Physicists
Credits: 3
Grad Meth: Reg, P-F, Aud
Prerequisite: PHYS165, and PHYS274 or MATH240, or permission of instructor.
Restriction: Permission of the CMNS-Physics department.
Survey relevant topics in contemporary machine learning (ML) to develop a conceptual understanding of important techniques and an ability to implement them in practice using python. Linear models: linear and logistic regression, support vector machines and kernel methods. Basic aspects of information theory and probability relevant for ML. Neural networks: architectures (FCN, CNN, RNN, attention and transformers) and initialization schemes (order-chaos transition, information propagation). Optimization algorithms. Neural tangent kernel, infinite limits of neural networks, neural scaling laws. Basic techniques in unsupervised learning including dimensionality reduction and generative models.