Interpretable Machine Learning

Spring 2019
Mini-1 T/H 3:00-4:20pm
Picture of Adam Perer
Adam Perer
(Instructor)

Machine learning has essentially become ubiquitous for making sense of large and complex data. While these techniques may be automated and yield high accuracy precision, they are often black-boxes that limit interpretability. Interpretability is acknowledged as a critical need for many applications of machine learning, and yet there is limited research to determine how interpretable a model is to humans.

This course will be is a reading group-style seminar, where each class we will discuss recent research papers on different aspects of interpretable machine learning. While the field and definition of interpretable machine learning varies by researchers and practitioners, this seminar will focus on interpretability techniques that are human-centered and aim to help people understand the ML model and its implications.

The goal of the course is to introduce students to current practices and future research opportunities about how to better enable the interpretability of machine learning.

Syllabus

Course Goals

Machine learning has essentially become ubiquitous for making sense of large and complex data. While these techniques may be automated and yield high accuracy precision, they are often black-boxes that limit interpretability. Interpretability is acknowledged as a critical need for many applications of machine learning, and yet there is limited research to determine how interpretable a model is to humans.

This course will be is a reading group-style seminar, where each class we will discuss recent research papers on different aspects of interpretable machine learning. While the field and definition of interpretable machine learning varies by researchers and practitioners, this seminar will focus on interpretability techniques that are human-centered and aim to help people understand the ML model and its implications.

The goal of the course is to introduce students to current practices and future research opportunities about how to better enable the interpretability of machine learning.

Work Required

Work required for this course is the following:

2-4 Readings per week. In-class Quizzes on Readings and Lecture Material. 1-2 times per semester: Lead and create a team presentation for a Research Paper. Participating in Panel Discussions Others Are Leading There is no final exam or final project in this course. Students who do well will be invited to continue on an independent project on topics related to the course, working with Prof. Perer during a future semester.

More About Research Paper Presentations Each class, we will have a Research Paper presentation by a team of 2 students in the class. Students will get to vote on the papers most interesting to them, and assigned to a team accordingly. This will allow students to focus on papers more relevant to their research interests and background (e.g. data visualization, deep learning, or fairness of ML)

Students team should create a 20-minute presentation describing the Research Paper. Students will create slides to summarize the research paper. The presentation should include examples of how the interpretability technique works, preferably using videos and/or demos made available by authors. The team should also be prepare to lead a panel discussion, with the help of Prof. Perer, by answering questions either directly from other students or the instructors, or pulled from the questions asked on the Slack channel that week.

Students not presenting in a given week will be responsible for asking a question in the Slack channel, and upvoting or downvoting at least one question from others.

Grading for Research Paper Presentations will be based on overall preparation and quality of the presentation, ability to answer questions with supporting material from the readings or elsewhere, and participating in asking questions on the Slack channel.

Grading

Final grades will be determined as follows:

  • 20% Quizzes
  • 50% Research Paper Presentations
  • 20% Research Paper Participation (on Slack and in class)
  • 10% Attendance

We intend for anyone who puts in the effort required to be able to achieve a B or better in this course. Quizzes, Research Paper Presentations, and participation are intended to be straightforward to complete and/or prepare (albeit require significant effort).

Missing Class

Since this course is a reading seminar style course, attendance is mandatory, as its critical for students to attend to fully participate in this course . However, I also realize that many students have obligations that may require them to miss an occasional class. Each student will receive 2 “free” classes to miss, without penalty. We will also drop the lowest two grades on quizzes. If you have a medical or other excused absence, please let me know as soon as possible, and I will work to accommodate you.

There will be no opportunity to make up in-class activities (quizzes, reading paper participation) beyond the two free drops.

Office Hours

Please join me for office hours. My normal office hours will be TBD in Newell-Simon Hall (second floor). Please also feel free to send a message to him on Slack.

If you are asking a general question, that other students may also benefit from seeing, please ask in the general Slack channel. If you are asking a question specific to you, /e.g./, about a grade you received, about absence from class, an accommodation request, etc., then please ask us individually either in person or via Slack DM.

We try to respond very quickly, but please do email me again if you don’t receive a response within 24 hours.

Accommodations for Students with Disabilities If you have a disability and have an accommodations letter from the Disability Resources office, I encourage you to discuss your accommodations and needs with me as early in the semester as possible. I will work with you to ensure that accommodations are provided as appropriate. If you suspect that you may have a disability and would benefit from accommodations but are not yet registered with the Office of Disability Resources, I encourage you to contact them at access@andrew.cmu.edu.

Health and Well-being

If you or anyone you know experiences any academic stress, difficult life events, or feelings like anxiety or depression, we strongly encourage you to seek support. Counseling and Psychological Services (CaPS) is here to help: call 412-268-2922 and visit their website at http://www.cmu.edu/counseling/. Consider reaching out to a friend, faculty or family member you trust for help getting connected to the support that can help. If you or someone you know is feeling suicidal or in danger of self-harm, call someone immediately, day or night: CaPS: 412-268-2922 Re:solve Crisis Network: 888-796-8226 If the situation is life threatening, call the police On campus: CMU Police: 412-268-2323 Off campus: 911

If you have questions about this or your coursework, please let me know. Thank you, and have a great semester.

(Some of this helpful content was borrowed from Prof. Jeff Bigham’s Human-AI Interaction Course)