Exams
Overview
Your grade is based on two components: a midterm exam and a final assessment. For the final, you choose between an exam or a project with presentation.
| Component | Points | Weight | Format |
|---|---|---|---|
| Midterm | 0-30 | 30% | Written, closed book, 1 hour |
| Final Exam | 0-30 | 70% | Coding (open book) + written, 2 hours |
| OR Final Project | 0-30 | 70% | Report + presentation |
Final grade = (Midterm / 30) × 30% + (Final / 30) × 70%
Students who choose the project path receive +3 bonus points on the final mark, which can push the grade to 30L.
Midterm Exam
Details: date, format, what to study, mocks
| Date | April 16, 2026 |
|---|---|
| Time & Room | Same time and room as regular lectures |
| Format | Written |
| Duration | 1 hour |
| Materials | Closed book, no notes, no electronics |
| Coverage | Sessions 01-05 (Parts 1-2: Foundations, and The Science of Experimentation and Uncertainty) |
| Points | 0-30 |
What to Expect
The midterm tests your conceptual understanding. You will be asked to:
- Explain core concepts in your own words
- Identify flaws in experimental designs or evaluation pipelines
- Interpret results and explain what they mean (and don't mean)
- Apply concepts to new scenarios
What to Study
The midterm covers material from the course reader (from Stanford's CS109): Parts 1–4, up to and including Bootstrapping.
Part of the midterm is also eval in ML — see the Methodology reading (a shortened version of material from the MLVU course at VU Amsterdam).
Exclusions — you are not responsible for:
- Bayesian Networks
- Any "Applications" sections
- The shape/plot of probability distributions, except Bernoulli, Binomial, and Normal — for all others, you are expected to know you can look them up and how to do so
Exam Structure
The midterm consists of computation exercises (like the mock exams below) plus a couple of short theory questions that ask you to explain a concept in your own words. For example: "Why, in assessing the quality of ML models, do we only look at the test dataset size, and not at the train vs test proportion?"
Pre-Exam Mocks
- Mock Midterm #1 — with solutions
- Mock Midterm #2 — with solutions
- Mock Midterm #3 — extra practice (no solutions)
Final Assessment
Part II: choose one option (70% of the overall exam mark)
For the remaining 70%, choose one of the following:
Option A (discouraged): Final Exam (0-30 points)
| Format | Coding + written answers |
|---|---|
| Duration | 2 hours |
| Materials | Open book, notes allowed, internet allowed (for the coding part). The written part is closed-book |
| Coverage | Full course, emphasis on Sessions 06-11 (Parts 3-5: Representations, Graphs, Conversational "X") |
| Setup | Bring laptop; Jupyter notebook provided |
Exam Structure
| Section | Weight | Description |
|---|---|---|
| Part A: Concepts | 30% | Short answer questions and exercises on similarity, clustering, graphs and other part II topics |
| Part B: Applied | 40% | Solve a problem: analyze a dataset, write code, produce results |
| Part C: Reporting & Discussion | 30% | Discuss your Part B solution: limitations, alternative approaches, what you would investigate next |
Note: Part C builds on your Part B work. Even if your code didn't fully work, you can earn full marks in Part C by discussing what you intended, why, and its limitations.
Pre-Exam Mock
- Mock Final — Part II — graphs / centrality + a knowledge discovery and analysis coding task
Option B (strongly encouraged): Final Project (0-30 points + 3 bonus)
A deeper dive into a topic of your choice — study-first, with a defensible empirical claim. See the Capstone projects page for the full brief, scaffolding, deliverables, and rubric.