CS 282A. Designing, Visualizing and Understanding Deep Neural Networks

Catalog Description: Deep Networks have revolutionized computer vision, language technology, robotics and control. They have growing impact in many other areas of science and engineering. They do not however, follow a closed or compact set of theoretical principles. In Yann Lecun's words they require "an interplay between intuitive insights, theoretical modeling, practical implementations, empirical studies, and scientific analyses." This course attempts to cover that ground.

Units: 4.0

Student Learning Outcomes: Students will learn design principles and best practices: design motifs that work well in particular domains, structure optimization and parameter optimization. , Students will come to understand visualizing deep networks. Exploring the training and use of deep networks with visualization tools., Understanding deep networks. Methods with formal guarantees: generative and adversarial models, tensor factorization.

Prerequisites: MATH 53 and MATH 54 or equivalent; COMPSCI 70 or STAT 134; COMPSCI 61B or equivalent; COMPSCI 189 or COMPSCI 289A (recommended)

Fall: 3.0 hours of lecture and 1.0 hours of discussion per week
Spring: 3.0 hours of lecture and 1.0 hours of discussion per week

Grading basis: letter

Final exam status: No final exam

Class Schedule (Spring 2022):
MoWe 5:00PM - 6:29PM, Dwinelle 155 – Jensen Gao, Marvin M Zhang, Xinyang Geng

Spring 2020 class homepage on bCourses

Class homepage on inst.eecs

General Catalog listing