Machine Learning (ECE), EECS 553, Fall 2022

The goal of machine learning is to develop computer algorithms that can learn from data or past experience to make accurate, useful, and fair predictions on new unseen data. In the past few decades, machine learning has grown tremendously, and it has made major impacts in many real-world applications. This course will give a graduate-level introduction of machine learning and provide mathematical/statistical foundations of machine learning, mathematical derivation and implementation of the algorithms, and discussion of their applications. On assignments, students will work through mathematical derivations applying principles learned in class, implement machine learning algorithms in Python, and apply those algorithms to data sets spanning a variety of applications. In small groups, students will also participate in the Fall 2022 Machine Learning Reproducibility challenge, which involves reproducing the results and assessing the conclusions of a paper published in a major machine learning conference.

Instructor: Professor Laura Balzano,
Course time: Tuesday/Thursday 1:30-2:50pm
Course location: GGBL2505. All lectures will be recorded.
Office hours: See Syllabus
GSI(s): See Syllabus
Textbook: None required, but there are several useful references that will be listed on Canvas.

Note for Thursday, September 15: You can find the current official syllabus here. We have recently revised the schedule and you can find information in the syllabus as well as on Canvas.

We have recently revised the schedule and you can find information in the syllabus as well as on Canvas.

The class waitlist is now empty and the class is not fully enrolled! If you enroll you will automatically get an override the next day.

Note for Friday, August 30: I sent an email to students a few days ago, which you can find here.

GGBL 2505, our new classroom, holds 130 students. As of August 30, there are 102 enrolled, 12 with permissions to enroll, 13 who will be given permissions soon, and 153 on the waitlist.

Note for Friday, July 22:

I sent an email to the enrolled and waitlisted students today, which you can find here.

As of July 22, there are 71 enrolled and 277 on the waitlist.

An update with the status of the waitlist, class size, and other details as well as new FAQs will be posted here at least once more before the start of the semester.

FAQ:

What is the difference between EECS 445, 453, 545 and 553?

Starting in Fall 2022, EECS 453/553 are offered by the ECE division. EECS 445/545 are offered by the CSE division. Note: EECS 453 is numbered EECS 498 for Fall 2022. Due to this recent new course numbering, things you find written online may be out of date.

Both 545 and 553 will assume familiarity with probability and linear algebra. EECS 553 specifies experience with graduate coursework in linear algebra and probability. The primary difference between the courses is a matter of emphasis. While both courses will cover fundamental principles, mathematical derivations, and applications to real-world examples, 445/545 will place relatively more emphasis on software implementation and application to real-world data, whereas 453/553 will place relatively more emphasis on mathematical derivations and principles. Both 500-level courses will serve students seeking to do research in machine learning.

Can I take the course even if I haven't had graduate coursework in probability and linear algebra?

I discourage students who have not taken probability and linear algebra courses at 500 level or above from taking EECS 553 in Fall 2022. I am unlikely to admit students without these pre-requisites from the waitlist. I encourage you to take EECS 498 - Principles of Machine Learning, or take the pre-requisite courses and then take 553 in future.

Will lectures be recorded?

Yes, all lectures will be recorded using the lecture capture system. Most classes will not be synchronously on zoom, but any class on zoom will be recorded as well.

Can I take 553 along with another class/lab whose classtime overlaps?

No.

There are a lot of people on the waitlist; will the class size be increased or new sections added?

There are always a lot of people on the waitlist for 545/553. The department/divisions do their best to offer as many sections as possible, but any given semester we may not have capacity. This semester no new sections will be added. If more GSIs are available and the classroom can be changed to a larger one, we may be able to increase the class size. Note that both EECS 545 and EECS 553 are slated to be taught in Winter 2023.

I am on the waitlist and would like to get an override, what do I do?

Fill out the override request form that is linked in an email you will receive / have received in the week of July 25. Include whether you have taken the necessary prerequisite courses. If the probability and linear algebra courses you took are not listed on my syllabus, please list their UM course numbers and/or describe the courses taken elsewhere. Then check your email daily to see if an override was issued. To the best of my knowledge, 553 is not strictly necessary for any degree program. If you believe your degree requires 553, in your override request please include the text of your degree requirements copied from the program manual, and please also include the URL of the program manual.

I received an override, what do I do?

Follow the instructions in the email promptly. See this video for instructions on how to use an override.

I am in spot N on the waitlist, will I get in?

See the priority order in the syllabus to make a guess at whether you'll get in. Since I am emphasizing the advisory prerequisite graduate coursework, there may be some people dropping from enrollment and the waitlist. That said, if you have taken those courses, don't lose hope; people drop all the way through the second week of class.