Homeworks
In this section, we will post the biweekly homeworks for the semester. Check back here for the latest problem set!
Homework 0
- Material
- Release
- Monday, January 12th, 7:30 PM ET
- Due
- Friday, January 23rd, 11:59 PM ET
Homework 1
- Material
- Release
- Tuesday, January 20th, 2:30 PM ET
- Due
- Tuesday, February 3rd, 11:59 PM ET
Update (01/25, 01/28): Several small updates have been made for clarity to Problem 3 and Problem 4: Problem 3(c): The previous version of PS1 had a line that said to “remember to divide by $n$” but that isn’t necessary to the problem, so we have removed it.
4(c): Previously, there was text in this problem asking you to generate a test set of size 1000 1000, but there was nothing to do with this test set in 4(c). We removed it to avoid confusion.
Problem 4(d): We clarified that the $d$ that you are adjusting in this problem is the one for your design matrix (while keeping $d = 5$ for your true function). We also noted that the “theoretical” behavior you might expect for this question for getting a perfect fit might not be exhibited in your code. That’s fine – you can write a one-sentence conjecture here for why that might be the case (anything is worth credit).
4(e): In this problem, it wasn’t explicitly clear how large of an independent test set to draw. We added text guidance to always draw 1000 1000 samples for your independent test set (separately from your training data for different values of n).
Homework 2
- Material
- Regularization and Gradient Descent
- Release
- Tuesday, February 3rd, 2:30 PM ET
- Due
- Tuesday, February 17th, 11:59 PM ET
Homework Submission Instructions
Submission for all the problem sets is handled through Gradescope. If you have not used Gradescope before, please watch this short video: “For students: submitting a PDF on the Gradescope website.”. A typical problem set will have several multi-part theoretical problems and a single coding problem to exercise the students’ ability to implement certain algorithms from scratch.
Collaboration Policy: Learning is best done in collaboration with peers. To this end, you will be allowed to collaborate with other students on the problem sets. All collaborators must write the names and UNIs of their group at the top of each problem set. All collaborators must also type up everything in their own words (and submit separately). You are free to discuss, whiteboard, and brainstorm with your collaborators. However, when it comes to sitting down and solving the actual problem, you must do it yourself, away from your collaborators.
LLM Policy: We strongly suggest resisting the temptation of using an LLM to help you solve your assignments at all. You will hopefully find that fluency on the homeworks will directly translate to exam performance, as the exam questions will track problem set questions closely. Of course, there’s no way for the instructor to really know whether you used an LLM. Due to this, please at least refrain from using LLMs to seek a verbatim answer on problem sets for your own good. If you must, you may query LLMs for prerequisite material you may have forgotten or to “explain” concepts in a simpler way and cite the LLM you used on your problem.
Drop Policy: The lowest problem set will be dropped for every student.
Late Policy: Students will have a total of 6 late days for all assignments. After the late days are used, late homeworks can be accepted for 48 hours after the time it is due, with a 20% penalty per day. The maximum late days acceptable for each assignment is two late days; Gradescope will not accept assignments 48 hours after the due date.
Submission Format: All submission will be through Gradescope. At the beginning of the semester, you will be added to the Gradescope roster. To submit an assignment, you will need to:
- Upload a single PDF document containing all the math, code, plots, and exposition required for each problem. You will not be uploading the code itself to Gradescope.
- Begin each problem on a new page.
- Select appropriate page ranges on Gradescope for each homework problem, as described in the “For students: submitting a PDF on the Gradescope website.” video.
It is strongly encouraged that you submit your homework PDFs as typesetted with LaTeX. Overleaf is a standard online platform for typesetting LaTeX, a bit like Google Docs. You are welcome to generate your LaTeX submissions however you’d like; we only need your PDF. Here’s a quick resource to get you up to speed for typesetting in LaTeX. A standard way to include Python code in your PDF generated by LaTeX is minted.
Another alternative is to submit your assignments as a Jupyter notebook converted into a PDF. The Markdown portions of Jupyter notebooks allow you to typeset math as well. As long as the submission is a single PDF file, the source that generates it is up to you. If you decide to submit this way, please make sure that the only outputs included in your submission PDF are from code cells that are relevant to the assignment.