Deep Networks have revolutionized computer vision, language technology, robotics and control.
They have a growing impact in many other areas of science and engineering, and increasingly,
on commerce and society. They do not however, fully follow any currently known 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 is a fancy way of saying “we don’t understand
this stuff nearly well enough, but we have no choice but to muddle through anyway.”
This course attempts to cover that ground and show students how to muddle through even
as we aspire to do more. That said, we will be leveraging the substantial, though still
tentative, understanding that we have gained in the past few years. It isn't 2015 anymore...
We know a lot more than we used to.
Lecture reference materials
Lectures are webcast by the department and recordings will be posted to a
youtube playlist.
You must be logged into your @berkeley.edu account to access the videos.
Lectures can have a substantial amount (if not all) of the content covered on the whiteboard or handwriting on a tablet (not on
presentation slides). Whether material is presented on slides or on the board, students are expected to take timely handwritten notes (before the next lecture or discussion occurs) and use those to study.
Because Deep Learning is rapidly evolving field, the material covered in this course can change substantially from semester to semester. Our goal is to bring you as close to the current frontier as we can --- while staying within what we understand most stably so that you have a solid foundation for the future.
If interested in materials from previous iterations of this course, please see here:
[Sp21],
[Fa22], [Sp23], [Sp25].
Do not refer to material from Fall 2024 since that is completely unvetted.
