Dr. James Anderson
Columbia University
Seminar Information
In this work I will discuss how bisimulation can be used as a unifying notion of task heterogeneity for multi-task learning in control, spanning classical LTI systems and modern latent world models. In the LQR setting, bisimulation-inspired metrics quantify heterogeneity across tasks and lead to explicit performance guarantees for multi-task and federated learning. I will then show how the same ideas improve robustness in Joint Embedding Predictive Architectures (JEPAs), where standard latent predictive models are often sensitive to control-irrelevant “slow features.” By incorporating a bisimulation encoder, we obtain compact latent representations that support robust model predictive control under significant visual variation. Together, these results highlight bisimulation as a principled tool for representation learning, generalization, and planning across a broad class of control problems.
James Anderson is an Associate Professor of Electrical Engineering at Columbia University where he is also a member of the Data Science Institute. From 2016 to 2019, he was a senior postdoctoral scholar in the Department of Computing + Mathematical Sciences at the California Institute of Technology. Prior to Caltech, he held a Junior Research Fellowship at St John’s College, University of Oxford, and was affiliated with the Department of Engineering Science. He received his DPhil (PhD) from Oxford in 2012 and his BSc and MSc degrees from the University of Reading in 2005 and 2006, respectively. His research spans control, learning theory, and optimization, with applications in smart grids and energy markets. Together with his students and collaborators, he has received several best paper awards in venues such as the IEEE Transactions on Control of Network Systems, the IEEE Conference on Decision and Control, and the Learning for Dynamics and Control Conference.