MLPR 2023 activities timeline
As soon as possible, work through the background section of the notes. The administration page explains how the course is assessed and how the activities work.
Links to question sheets and activities will appear throughout the Semester. The precise course content may change. You may need to refresh the page.
Weekly checklist:
- In-person lectures (recordings made available) (9am–9:50am Tue/Wed/Thu, weeks 1–10).
- Weeks 1–10: Lectures: 9am–9:50am Tue/Wed/Thu:
- Tue/Wed: Appleton Tower, Lecture Theatre 3
- Thursday: Appleton Tower, Lecture Theatre 1.
- Lecture Recordings (within 24 hours, need class registration).
- No lectures/recordings in week 2. See videos in notes w2a–w2e.
- Weeks 1–10: study the notes, including in-note questions (not assessed).
- Weeks 2–8: prepare for the next week’s tutorial.
- Weeks 3–9: Tutorials. (See administration page for details.)
- Weeks 3–9: assignment work as appropriate.
Week 1: 18 September –
- First class meeting: Tuesday 19 September, 9am.
- Sign up for the class Forum (hypothesis).
- w1 notes: Introduction to ML with Linear Regression
- Work through any remaining background material.
Week 2: 25 September –
- Watch the videos in notes w2a–w2e instead of live lectures this week.
- w2 notes: ML fundamentals: generalization, error bars, Gaussians
- Reminder: studying the notes includes doing the questions; don't fall behind from the start! Check your status.
- Prepare for Tutorial 1 next week. (Tutorial administration information.)
Week 3: 2 October –
- w3 notes: Classification and gradient-based fitting.
- Tutorial 1 questions (for Thu or Fri).
- Tell us your assignment 1 pair preference. First allocation of pairs at 2pm Tuesday (3 Oct). We need your pair (or no pair) preference before you can enter answers. Please be patient while we process these.
- Start work on Assignment 1 (print). You can do it by yourself or in a pair.
- Prepare for Tutorial 2 next week.
Week 4: 9 October –
- w4 notes: Bayesian linear regression
- Tutorial 2 questions (for Thu or Fri).
- Prepare for Tutorial 3 next week.
- Finish Assignment 1 (print).
Week 5: 16 October –
- Assignment 1 due Monday 16 October, 12 noon. (0% of final mark)
Feedback is only guaranteed for submissions made on time. - w5 notes: Bayesian model choice and Gaussian processes
- Tutorial 3 questions (for Thu or Fri).
- Prepare for Tutorial 4 next week.
- Tell us your assignment 2 pair preference.
Week 6: 23 October –
- w6 notes: More detailed models: Gaussian process kernels, more non-Gaussian regression
- Tutorial 4 questions (for Thu or Fri).
- Prepare for Tutorial 5 next week.
- Look at Assignment 2 (print) and schedule when you'll do it in weeks 6–9. You can do it by yourself or in a pair.
- Tell us your assignment 2 pair preference by 2pm Tuesday (24 Oct).
Week 7: 30 October –
- w7 notes: Neural Networks
- Tutorial 5 questions (for Thu or Fri).
- Prepare for Tutorial 6 next week.
Week 8: 6 November –
- w8 notes: Autoencoders, PCA, Netflix Prize
- Tutorial 6 questions (for Thu or Fri).
- We hope you haven't forgotten Assignment 2 (print).
- Prepare for the final Tutorial 7 next week.
Week 9: 13 November –
- w9 notes: Bayesian logistic regression, Laplace approximation
- Tutorial 7 (last one).
- Finish Assignment 2 (print), if you haven't already.
Week 10: 20 November –
- Assignment 2 due Monday 20 November, 12 noon. (25% of final mark)
Late work and extensions subject to Rule 1 of the School Late Policy. - w10 notes: Sampling-based approximate Bayesian inference, variational inference
- (No tutorial)
- Please fill out the course survey by 21 December.
- Take a few days off! Then revise for the exam.
December Exam (75% of your final mark)
- The date and time of the December 2023 exam is available on the Exam timetable (search for MLPR's course code INFR11130 on this form).
- We expect the exam to be a closed-book pen-and-paper exam, held in person in Edinburgh. However this is still subject to confirmation by the University.
- The library has past exam papers. Although you may not be able to do every question: the course has changed slowly over the years, after a big change in 2016. There is no past paper from 2020.