Full Stack Deep Learning
UW CSEP 590C Spring 2020
Instructor: Sergey Karayev
TAs: Shivan Singhal and Zhaofeng Wu
Our course will be entirely online.
We will meet 6:30 - 9:20 pm every Thursday according to the schedule below.
If you want to meet for 30 min “office hours”, please Slack me!
All synchronous meetings will occur on Zoom and will be recorded. Make sure you install the client beforehand, and test your audio/video so that you can participate in discussions. A decent webcam and microphone or headset greatly improve everyone’s experience.
All asynchronous communication will be on Slack (you will be added to the workspace). Although Slack is a type of chat, it is best used in a more asynchronous way. Use threads, don’t expect immediate responses, and write your messages to be more like emails and less like texts. Course staff will be most active on Thursdays, but our real goal here is to foster student discussion and community.
There is an official mailing list for this course, but it will not be used. Please communicate with us and each other via Slack.
Lastly, we’re living through some pretty stressful times right now. I understand that and want this class to be a source of support and comfort, not stress. I came across the slide below, and think it’s a valuable message to hear.
If there is a suggested reading, please read it after the lecture and before submitting the assignment.
Weekly assignment will be released 9pm Thursday, and due 6pm the following Thursday. No late submission will be allowed. All assignments, including the final exam, will be on Gradescope (you will be added to the course).
The final grade will be calculated as weighing each of the 8 assignments by 8%, and the final exam by 36%. Grading will be generous, because we’re all adults here.
In addition, we will have 8 labs, all ungraded, in which we will build a whole deep learning project to understand the content of handwritten paragraphs. We will walk through each lab together synchronously, and you are encouraged to spend as much time as you want on the labs independently.
Our computing environment for the labs will be web-based. You don’t need to have access to a GPU or set your machine up, beyond being able to access the Internet.
This course was originally developed as a 3-day weekend bootcamp, together with Pieter Abbeel and Josh Tobin. If curious, you can preview lectures by watching recordings from a year ago on that site.
Distill.pub has great articles filled with unprecedented visualization of how and why deep learning works.
Fast.ai is a great free two-course sequence aimed at first getting hackers to train state-of-the-art models as quickly as possible, and only afterward delving into how things work under the hood. Highly recommended for anyone.
/r/MachineLearning/ is a good community for staying up to date with developments.
The best deep learning newsletter in my opinion is The Batch by Andrew Ng.