Advanced Topics in Machine Learning: 2021-2022
Lecturers | |
Degrees | Schedule C1 (CS&P) — Computer Science and Philosophy Schedule C1 — Computer Science |
Term | Hilary Term 2022 (18 lectures) |
Overview
This is an advanced course on machine learning, focusing on recent advances in machine learning with relational data and on Bayesian approaches to machine learning. The course is organized and taught as follows:
- Relational learning (Ismail Ilkan Ceylan): 9 lectures
- Bayesian machine learning (Jiarui Gan): 5 lectures + Bayesian deep learning (Yarin Gal): 4 lectures
Learning outcomes
After studying this course, students will:
- Have knowledge of the different paradigms for performing machine learning and appreciate when different approaches will be more or less appropriate.
- Understand the definition of a range of neural network models, including graph neural networks.
- Be able to derive and implement optimisation algorithms for these models.
- Understand the foundations of the Bayesian approach to machine learning.
- Be able to construct Bayesian models for data and apply computational techniques to draw inferences from them.
- Have an understanding of how to choose a model to describe a particular type of data.
- Know how to evaluate a learned model in practice.
- Understand the mathematics necessary for constructing novel machine learning solutions.
- Be able to design and implement various machine learning algorithms in a range of real-world applications.
Prerequisites
Required background knowledge includes probability theory, linear algebra, continuous mathematics, multivariate calculus, and a basic understanding of graph theory, and logic. Students are required to have taken a Machine Learning course. Good programming skills are needed, and lecture examples and practicals will be given mainly in Python and PyTorch.
Synopsis
Relational Learning: Lectures 1–9, İsmail İlkan Ceylan
- Lecture 1. Relational data & node embeddings
- Lecture 2. Knowledge graph embedding models
- Lecture 3. Graph neural networks
- Lecture 4. Message passing neural network architectures
- Lecture 5. Expressive power of message passing neural networks
- Lecture 6. Higher-order graph neural networks
- Lecture 7. Message passing neural networks: unique features and randomization
- Lecture 8. Generative graph neural networks
- Lecture 9. Overview of applications of graph neural networks
Bayesian Machine Learning: Lectures 10–18, Jiarui Gan and Yarin Gal
- Lecture 10. Machine learning paradigms
- Lecture 11. Bayesian modeling 1
- Lecture 12. Bayesian modeling 2
- Lecture 13. Bayesian inference 1
- Lecture 14. Bayesian inference 2
- Lecture 15 & 16. Bayesian deep learning
- Lecture 17 & 18. Bayesian deep learning
Syllabus
Overview of relational learning and reasoning. Embedding models and knowledge graphs, inductive capacity of embedding models, graph representation learning, graph neural networks, expressive power of message passing neural networks, limitations and extensions. Overview of the Bayesian paradigm and its use in machine learning. Generative models, Bayesian inference, Monte Carlo methods, variational inference, probabilistic programming, model selection and learning, amortized inference, deep generative models, variational autoencoders.
Reading list
- William L. Hamilton. (2020). Graph Representation Learning.
Synthesis Lectures on Artificial Intelligence and Machine Learning, Vol. 14,
No. 3 , Pages 1-159. https://www.cs.mcgill.ca/~wlh/grl_book/ - Christopher M. Bishop, “Pattern Recognition and Machine Learning”, Springer, 2006
- Marc Peter Deisenroth, A. Aldo Faisal, and Cheng Soon Ong, “Mathematics for Machine Learning”, Cambridge University Press, 2020 https://mml-book.github.io/
Related research
Themes |