MASC 520: Mathematical Methods for Deep Learning (Spring 2023)

Instructor: Prof. Paulo Branicio

Course Summary: This is a foundational mathematical course for deep learning. It provides graduate students with in-depth knowledge of mathematics needed to understand deep learning. The course covers a variety of topics such as linear algebra, probability and statistics, optimization, Fourier series, Fourier transforms, ordinary and partial differential equations, and Markov Chain Monte Carlo methods. Each topic is introduced with an application of deep learning to problems in the physical sciences and engineering. Students are required to do five projects chosen from the following topics: feed-forward neural network, convolutional neural network, recurrent neural network, neural network solvers for differential equations, autoencoders, restricted Boltzmann machine and deep Boltzmann machine.