MATH 382 Mathematical Foundations of Deep Learning and Modern Extensions
This course introduces the mathematical foundations of deep learning, emphasizing modern models. Students study the structure of neural network architectures, including multilayer perceptron, convolutional neural networks, and transformers, with a focus on loss functions, gradient-based optimization, and training algorithms. Additional topics include self-supervised learning, reinforcement learning, and generative models such as variational autoencoders and diffusion models. Theory is reinforced through problem sets, hands-on labs, and readings from seminal research papers.
Prerequisite
Math 272 and Math 335, or permission of instructor
Offered
FallInstructor
Qin Lu