CSE 240D: Accelerator Design for Deep Learning (Fall 2019)


Instructor: Hadi Esmaeilzadeh


Email: hadi [AT] eng [DOT] ucsd [DOT] edu
Office: CSE 3228

TA:
FatemehSadat Mireshghallah fmireshg [AT] eng [DOT] ucsd [DOT] edu
Office hours: Thursdays 8-9AM, B240A

Course Description

Deep learning is set to revolutionize medicine, robotics, commerce, transportation and numerous other aspects of our lives. However, these impacts are contingent upon providing high-performance compute capabilities alongside restrained power consumption. Significant effort has been made both by modifying hardware and software in the direction of enhancing the speed of neural networks and their consumed energy. The compute intensity of neural networks and the inability of general purpose processors to meet this high demand accentuate the need for applications specific hardware accelerators that are custom designed for deep neural network computations. In this course, you will gain insight on the design process of these accelerators, as well as deep neural network architectures and characteristics by discussing the prevalent literature in the area. This is a project-based course, so you will also acquire hands-on knowledge on how to actually construct an accelerator through the project. The project will step by step guide you through outlining your architecture and developing a functional and timing simulator for it.

Format

The course will be a combination of lectures, student presentation, project and separate brainstorming sessions in which we will collectively explore and develop new ideas on each topic. The students will do a project in which they design an accelerator for deep neural networks and develop a simulator for their design.

Prerequisites

There are no pre-requisites for the course but students will be expected to be comfortable with a) basic computer organization, and b) programming in a language such as Python or C/C++. The course is open to both PhD and MS students. Undergraduate students require permission from the instructor.

Evaluation

Students will be evaluated based on the following rubric. For the details on each item please refer to its corresponding page.

Piazza Link:

piazza.com/ucsd/fall2019/cse240d

Class Participation 10%
Critiques 15%
Class Presentations 25%
Final Project 50%

Course Material

There is no required textbook. All relevant materials will be made available online.

Academic Honesty

Students are expected to abide by the UC San Diego Honor Code. Honest and ethical behavior is expected at all times. All incidents of suspected dishonesty will be reported to and handled by the office of student affairs. You will have to do all assignments individually unless explicitly told otherwise. You may discuss it with classmates but you may not copy any solution (or any part of a solution).