Machine Learning for Computer Systems
This offering runs for three weeks (Aug 31 – Sep 18, 2026) at the University of Chicago Center in Paris, twelve meetings instead of the usual eighteen. The syllabus describes the standard nine-week course; the items below override it for this term. Everything else in the syllabus stands.
| Component | This term |
|---|---|
| Exam | One combined exam on Thu Sep 17, in class. No midterm. Designed for about 45–50 minutes; the full class period is available and there is no hard cutoff. Closed book except one 8.5×11 handwritten sheet, both sides. |
| Assignments | Two, not 3–5: Assignment 1 (Video Quality Inference), due Mon Sep 7, and the project proposal. No third assignment (class vote, Sep 10). |
| Project | Groups of three (four allowed with a written division of labor). One-page proposal due end of week 1 (Fri Sep 4), committed to the team’s repo. Show-and-tell Wed Sep 16: informal, go around the room, slides optional. Due Sun Sep 20, 11:59 pm Chicago time. It’s Git: you may keep pushing after the deadline and I will look at updates, up to about one day before the grade-submission deadline. See project. |
| Participation | Show up, be engaged. No reading responses and no in-class quizzes this term. |
| Late policy | Unchanged: 96 late hours for the term, tracked from commit timestamps, no need to ask. |
Submission is by Git: fill in the intake form (Canvas) with your GitHub username and repo URL; commit assignment notebooks, the proposal, and the project to that repo. Scores come back by Slack DM; written feedback arrives as GitHub issues on your repo.
The exam is drafted from the class agenda and past exams with the prompt in
prompts/generate-summer-final.md. You have all of the inputs, so you can
generate your own practice exams the same way: point the prompt at
agenda/2026-summer.md and the past midterms
and finals.
The schedule table on the course page lists the standard lecture order. This term the meetings went:
| Meeting | Date | Lectures (schedule numbers) | Hands-on |
|---|---|---|---|
| 1 | Mon Aug 31 | 1 Introduction; syllabus | 01 Packet capture |
| 2 | Tue Sep 1 | 2 Security | 02 Scanning |
| 3 | Wed Sep 2 | 3 Performance, 4 Resource optimization | 03 QoE inference |
| 4 | Thu Sep 3 | 6 Feature extraction (incl. nPrint), start of 7 | 06 netml features |
| — | Fri Sep 4 | Excursion: Inria Nancy (no class) | |
| 5 | Mon Sep 7 | 8 Model training and evaluation | 08 Full pipeline (HTTP/log4j) |
| — | Tue Sep 8 | Excursion: Huawei France HQ (no class) | |
| 6 | Wed Sep 9 | 10 Linear regression, 11 Logistic regression, 12 Trees and ensembles | 10 Linear regression + basis expansion |
| 7 | Thu Sep 10 | 13 Deep learning (with 14 nPrint recap) | 12 IoT trees/ensembles, 13 DDoS neural net |
| — | Fri Sep 11 | Excursion: ENS de Lyon (no class) | |
| 8 | Mon Sep 14 | 15 Dimensionality reduction (incl. autoencoders), 16 Clustering | 15 PCA, 16 k-means |
| 9 | Tue Sep 15 | Exam review walk-through; 17 Diffusion / NetDiffusion, 19 Transformers and state-space models | 18 NetSSM walkthrough (Colab, GPU) — link fixed on the course page |
| 10 | Wed Sep 16 | Project show-and-tell; 20 Timeseries; 23 Model Performance and Maintenance (new deck: LEAF, CATO, AC-DC, ServeFlow) | NetSSM Colab |
| 11–12 | Sep 17–18 | Exam (Thu, proctored by Samuel), wrap-up |
Skipped or merged this term, and therefore not on the exam: 5 Data acquisition (covered only in passing), 9 Naive Bayes (the spam example in lecture 2 is fair game), the SVM half of 11, the separate nPrint hands-on (14), and the standalone autoencoder lecture (22; the autoencoder material in lecture 15 is fair game). Lecture 18 (state-space models) was covered inside Meeting 9’s generative-models discussion. Bonus lectures 20 (Timeseries) and 23 (Model Performance and Maintenance) were given on Wed Sep 16 after the exam draft was set; they are not on the exam. 21 and 22 were not covered.
Three course excursions, organized by the John W. Boyer Center in Paris with a student coordinator on each. Bring government ID; the hosts check names against the participant list.
| Date | Where | What |
|---|---|---|
| Fri Sep 4 | Inria Nancy – Grand Est / LORIA, Université de Lorraine, Nancy | 6:00 am pickup at the Mease residence for the morning train from Gare de l’Est. Presentation and tour of the Inria technology-development hub; meetings with the center’s director (Isabelle Chrisment) and several research and development teams; lunch hosted by the university; guided walking tour of historic Nancy; 4:24 pm train back to Paris (arrives 6:03 pm). |
| Tue Sep 8 | Huawei France headquarters, Boulogne-Billancourt | 8:40 am pickup at the Mease house; visit from about 10:00 am to 12:30 pm, hosted by Dario Rossi. Afternoon free; class was not held that morning. |
| Fri Sep 11 | ENS de Lyon, Lyon | 6:30 am pickup at the Mease residence for the TGV. Talk and lab presentation by Francesco Bronzino (co-author of the video QoE assignment) and colleagues, campus tour, lunch, walking tour of Vieux Lyon; evening train back (arrives 8:10 pm). |