Theoretical Machine Learning (Fall 2026)

Notes

• Each homework is worth 10% of the total grade (the entire homework is 40%).

• Solutions must be written as a PDF. The standard way to do so is to write your solutions in LaTeX and then compile them into a PDF (LaTeX source files will be provided; feel free to use them or any other templates you like). In the age of AI, however, you can also use AI to covert your handwritten solutions into a PDF if you think that is more efficient. Regardless of how the PDF is produced, you are responsible for ensuring that it accurately reflects your intended solutions; any errors introduced during AI-assisted conversion are your responsibility.

• Collaboration is allowed, so is using AI tools. However, the primary purpose of the homework is to practice solving theoretical problems and developing proofs independently. You are strongly encouraged to make a serious attempt at each problem before seeking substantial help from your classmates or AI. You are responsible for understanding and verifying everything you submit, regardless of how you obtained it. Short in-class quizzes will test your understanding of the ideas and techniques involved in the homework. Ultimately, use collaboration and AI as a tool to help you learn, not as a substitute for learning.

• Submit one single PDF through Gradescope.

• Late submission policy: you are given a total of 3 late days for the 4 problem sets, and you can use at most 2 late days for each homework. Additional late days will each result in a deduction of 10% of the grade of the corresponding assignment.