[edit]

ML and the Physical World

Welcome to the course Machine Learning and the Physical World. The course is focused on machine learning systems that interact directly with the real world. Building artificial systems that interact with the physical world have significantly different challenges compared to the purely digital domain. In the real world data is scares, often uncertain and decisions can have costly and irreversible consequences. However, we also have the benefit of centuries of scientific knowledge that we can draw from. This module will provide the methodological background to machine learning applied in this scenario. We will study how we can build models with a principled treatment of uncertainty, allowing us to leverage prior knowledge and provide decisions that can be interrogated.

Lectures

Week 1 Session 1: Introduction: ML and the Physical World

Lecturer: Neil D. Lawrence
Date: Friday, October 08, 2021
Time: 12:00

Week 1 Session 2: Quantification of Beliefs

Lecturer: Carl Henrik Ek
Date: Tuesday, October 12, 2021
Time: 10:00

Week 2 Session 1: Gaussian processes

Lecturer: Carl Henrik Ek
Date: Friday, October 15, 2021
Time: 12:00

Week 2 Session 2: Simulation

Lecturer: Neil D. Lawrence
Date: Tuesday, October 19, 2021
Time: 10:00

Week 3 Session 1: Emulation

Lecturer: Neil D. Lawrence
Date: Friday, October 22, 2021
Time: 12:00

Week 4 Session 1: Probabilistic Numerics

Lecturer: Carl Henrik Ek
Date: Friday, October 29, 2021
Time: 12:00

Week 4 Session 2: Emukit and Experimental Design

Lecturer: Neil D. Lawrence
Date: Tuesday, November 02, 2021
Time: 10:00

Week 5 Session 1: Sensitivity Analysis

Lecturer: Neil D. Lawrence
Date: Friday, November 05, 2021
Time: 12:00

Week 5 Session 2: Multifidelity Modelling

Lecturer: Neil D. Lawrence
Date: Tuesday, November 09, 2021
Time: 10:00

Week 6 Session 2: Projects

Lecturer: Carl Henrik Ek
Date: Saturday, November 13, 2021
Time: 12:00

Case study Lectures

The lectures for weeks 5 to 8 will focus on “special topics”. The lecture on Thursday will be given by a guest lecture discussing a specific real world scenario where they have applied machine learning or machine learning would be applicable. During the Friday lecture Neil and Carl Henrik will lead a discussion relating the material back to the taught part of the course and provide an outlook of potential challenges that could provide the basis for a project. We will also use this time to discuss questions that have come up during the week.

Week 6 Session 1: Simulating Contact Tracing in the Pandemic: TTI Explorer

Lecturer: Andrei PaleyesBryn Elesedy
Venue: Virtual (Zoom)
Date: Thursday, November 12, 2020
Time: 10:00

Week 7 Session 1: Regression, Causality, Statistical Paradoxes and other Fairy Tales

Lecturer: Javier Gonzales
Venue: Virtual (Zoom)
Date: Thursday, November 19, 2020
Time: 10:00

Week 8 Session 1: Climate Science Challenges and Opportunities for Machine Learning

Lecturer: Scott Hosking
Venue: Virtual (Zoom)
Date: Thursday, November 26, 2020
Time: 10:00

Communication

The course will be run completely virtually, to make the most of this setting we are testing out a few new ideas. The first one will be that we will run a reddit feed for the course. The aim here is to create a forum for discussion and a public way to answer questions to the benefit of all. Importantly, the feed will be open to anyone and you can be completely anonymous while you will be able to identify us. The reason that we have chosen reddit as a forum is the self moderating structure. The idea is that we are all responsible for “upvoting” questions and comments we find interesting. This will allow us to focus on things that are of general interest. Furthermore we hope that the relaxed attitude of reddit will stimulate an open and opinionated discussion around anything that relates to the course. We will also use the discussion on the forum to inform us to better align the lectures to your understanding and during the last block to inform the Q&A session. We consider opinions and critique in all its forms positive, it evidence that you care.