Probabilistic Machine Learning
Prerequistites.
The content in this course is advanced machine learning. If you are looking for an introduction to probabilistic machine learning, please see the IB Data Science course.
Auditing & practical arrangements.
Students take the first two topics, and then their choice of topics 3 or 4.
- If you wish to audit this course, all the lecture material is available online, as listed below, and there is no need to ask permission to use it.
- For students registered on this course: coursework submission, Q&A, office hours, and briefing sessions are on Moodle.
Topics covered
1. Gaussian processes
-
Delivered by the Engineering department, the first part of
engineering course 4f13
Syllabus and slides, videos on Moodle, and Moodle enrolment key - Exercise 1 [notebook]
2. TrueSkill — Graphical models and Gibbs sampling
- Delivered by the Engineering department, as above
- Exercise 2 [notebook]
3. Probabilistic neural networks
These lecture notes are the advanced parts of a draft textbook. The background material in the rest of the textbook may be helpful, and you can find videos as part of IB Data Science.
-
3.1 Deep learning and
3.2 Probabilistic deep learning — video (20:27) - 3.3 Numerical optimization with Pytorch
- 3.4 Deep generative models — video (10:40)
- 10.6 Recurrent neural networks — video (9:09)
-
5.3 Importance sampling —
video (11:17)
→ see also section 5.1 on Monte Carlo estimation, from the textbook — video (13:15) - 5.4 Probabilistic autoencoders — video (28:37)
-
5.5 Deriving the autoencoder —
video (15:15)
→ see also section 4.7 on probability bounds, from the textbook - 9. Empirical validation of supervised and unsupervised models (no written notes) — video (7:40)
- Code: nn.ipynb
- Exercise 3 [notebook]
4. Models for document collections — LDA
- Delivered by the Engineering department, as above
- Exercise 4 [notebook]