Lecture 11 - Summary and Guest lecture

Content: Summary and guest lecture
Lecturer: Niklas Wahlström (summary) and Lawrence Murray (guest lecture)
Slides: Download lecture11_handout.pdf


Place: Zoom

The first half the lecture (15:15-16:00) will be a summary of the whole course and the second half starting at 16.15 will be a guest lecture. We are very happy to announce that the guest lecture is given by Lawrence Murray Links to an external site..

Lawrence is a senior research scientist at Uber AI Links to an external site., with interests including computational statistics, machine learning, probabilistic programming, and high-performance computing. He is the developer of the probabilistic programming languages Birch Links to an external site. and LibBi Links to an external site.. He has previously worked at Uppsala Links to an external site., Oxford Links to an external site. and CSIRO Links to an external site., holds a Ph.D. (informatics) from Edinburgh Links to an external site. and honours degree (software engineering) from ANU Links to an external site..

Title: Programmatic Models and Probabilistic Programming

Abstract: Just as we have graphical models---probabilistic models represented as graphs---we can think about programmatic models---probabilistic models represented as computer programs. In this lecture we'll look at programmatic models and their representation in the probabilistic programming language Birch Links to an external site.. We'll especially look at three critical tools that a probabilistic programming language provides over a regular programming language: automatic marginalization, automatic conditioning, and automatic differentiation, and how these combine into sophisticated inference algorithms in the framework of Sequential Monte Carlo.