Advances in Data Science and AI: Welcome Talks

In this new seminar sub-series, our new academic colleagues working in data science and AI-related areas introduce themselves to the IDSAI community with short talks on their current research.

At this first event, we will welcome Mauricio A Álvarez and Wei Pan. This will be a hybrid event, taking place at The University of Manchester and online.

Gaussian processes (GPs) are probabilistic models used in machine learning for supervised and unsupervised learning. Mathematically, Gaussian processes are prior distributions over function spaces and combined with Bayesian inference, they allow for uncertainty quantification in prediction problems. Multi-task Gaussian processes generalise GPs to the setting of several simultaneous prediction problems and make them suitable for settings like multi-fidelity learning and transfer learning. In this talk, I will briefly introduce multi-task GPs and review how we have used multi-task GPs for tasks with heterogeneous nature (e.g. tasks that are combinations of different types of regression) and tasks with different resolution (e.g. aggregated datasets).

Bio: Mauricio A Álvarez is a Senior Lecturer in Machine Learning in the Department of Computer Science and a member of the Centre for AI Fundamentals at the University of Manchester . Previous to joining Manchester, he was Associate Professor at Universidad Tecnológica de Pereira, Colombia, and Senior Lecturer at the University of Sheffield. He is internationally known for his work on multi-output Gaussian processes and physically inspired probabilistic modelling. Dr Álvarez is Associate Editor for the Statistics and Computing journal and the Transactions on Machine Learning OpenReview journal. He has been area chair for several machine learning conferences including the Advances in Neural Information Processing Systems (NeurIPS) conference, the Uncertainty in Artificial Intelligence (UAI) conference, the International Conference on Learning Representations (ICLR), the AAAI Association conference and the Artificial Intelligence and Statistics (AISTATS) conference.

Recent years have seen a great number of contributions to machine learning methods for real-world robotic control from both the control and machine learning communities. In many applications, how to formally certify the safety of a data- and learning-based control policy is paramount. We present a Bayesian method in reinforcement learning by modeling the control Lyapunov function and control barrier function in control theory as priors. The sporadically used Bayesian method shows a great promise in control performance, stability and safety guarantee, and computational cost. We will illustrate these “safe Bayesian reinforcement learning” methods on both simulated robot control tasks and real-world experiments using wheels robots, quadrotors, spacecraft.

Bio: Wei Pan is a Senior Lecturer in Machine Learning at the Department of Computer Science and as part of Centre for AI Fundamentals and Centre for Robotics and AI, The University of Manchester, UK. Before that, he was an Assistant Professor in Robot Dynamics at the Department of Cognitive Robotics and co-director of Delft SELF AI Lab, TU Delft, Netherlands and a Project Leader at DJI, China. He is the recipient of Dorothy Hodgkin’s Postgraduate Awards, Microsoft Research Ph.D. Scholarship and Chinese Government Award for Outstanding Students Abroad. He is an area chair or associate editor of CoRL, IEEE Robotics and Automation Letters, ICRA, IROS and AAAI. He received his degrees from Imperial College London, University of Science and Technology of China and Harbin Institute of Technology.

This event has been organised by the Institute for data Science and AI.

IDSAI is one of The University of Manchester's Digital Futures network themes.

Event Venue & Nearby Stays

A202, Samuel Alexander Building, Oxford Road, Manchester, United Kingdom