Topic: Uncertainty Quantification in Machine Learning
Recently, machine learning has seen enormous success in solving a broad range of real-world applications, from personalized recommendations in e-commerce platforms to materials design. In the near future, machine learning is expected to take over higher-stakes applications such as personalized medicine and extreme-weather forecasting. However, to achieve this, machine learning systems must be reliable. This, in turn, requires characterizing and estimating uncertainties in the predictions these systems make, which is challenging due to computational limitations and the roughness of real-world data. This course will study recent advances in the interplay between machine learning and uncertainty quantification aiming to address these challenges.
The goal of this course is for students to be able to:
Understand the interplay between machine learning and uncertainty quantification.
Recognize existing methods and potential application areas.
Effectively apply existing methods in pre-defined settings.
Explore new techniques and develop methods that work on real-world data.
Instructor and Teaching Assistants
Raul Astudillo (rastudil@caltech.edu)
Chris Yeh (cyeh@caltech.edu)
Amy Li (kli5@caltech.edu)
James Bowden (jbowden@caltech.edu)
Grading
Grades will be based on the final project (80%) and 4-5 assignments (20%).
Final Project
The final project may be done in groups (recommended group size is 2-3, max is 4); one final project document should be submitted per group. The deliverables are a project proposal due on May 8, and the final project report due on June 8. Students will present their projects in a poster session on June 6.
Assignments
Each student should submit a copy of each assignment, but students may work collaboratively. All students should understand all parts of the completed assignment.
Platforms
We will use Piazza, Slack, and Gradescope. Please use Piazza to submit assignment-related questions so that instructors can respond promptly and all students have access to the response. Please do not use Slack to ask questions that may be of general interest. Slack is meant to be used mainly for discussions related to research projects. Completed assignments and final projects should be submitted on Gradescope.
Late Policy
Each student has a total of 48 late hours for the term for all the assignments combined. There is no late submission allowed for the final project proposal and report.
Reading Material
Additional references are available here. It is not necessary to read all material in depth for this course - they are compiled to be a resource for students to explore additional topics relevant to the class, which may be useful for the final project and beyond.
Class Calendar