Keynote
Prof. Chris Williams
09:30 27 July 2022
University of Edinburgh | Professor of Machine Learning at the School of Informatics | Website
Previous | Next
Assisting Scientific Discovery with AI
In the first part of the talk I will discuss the processes of scientific discovery (hypothesis-driven and discovery-based), and where and how AI can be of assistance. I will also discuss how this can be supported in universities. In the second part of the talk I will discuss work on Multi-task Dynamical Systems (joint work with Alex Bird and Chris Hawthorne). We are interested in how time series models can be specialized to individual sequences while retaining statistical power by sharing commonalities across the sequences. We describe the multi-task dynamical system (MTDS), a general methodology for extending multi-task learning to time series models. Our approach endows dynamical systems with a set of hierarchical latent variables which can modulate all model parameters. We apply the MTDS to motion-capture data of people walking in various styles using a multi-task recurrent neural network (RNN), and to patient drug-response data using a multi-task pharmacodynamic model.