Philipp Schwartenbeck

I am interested in generalisable representation learning in biological and artificial brains. Specifically, I focus on problems where agents learn abstract task knowledge that can be generalised (re-used) within and across tasks. To probe such representations in biological brains I use functional magnetic resonance imaging (fMRI) and magnetoencephalography (MEG). A key focus is the idea that abstract knowledge might be instantiated via a ‘cognitive map’ in the medial temporal lobe and prefrontal cortex, and that neural ‘replay’ might play a central role in building and employing such knowledge. I also work on probabilistic models of representation learning and their link to reinforcement learning algorithms. These models make testable predictions for mechanisms of representation learning and their temporal evolution in biological brains, and are equally informed by the biological underpinnings of these processes.

Publications most reflective of my research focus are

Schwartenbeck P., Baram A., Liu Y., Mark S., Muller T., Dolan R., Botvinick M., Kurth-Nelson Z., and Behrens T. Generative replay for compositional visual understanding in the prefrontal-hippocampal circuit. bioRxiv 2021

Nour M., Dahoun T., Schwartenbeck P., Adams R., FitzGerald T., Coello C., Wall M., Dolan R., and Howes O. Dopaminergic basis for signaling belief updates, but not surprise, and the link to paranoia. Proceedings of the National Academy of Science 2018.

Schwartenbeck* P., FitzGerald* T., and Dolan R. Neural signals encoding shifts in beliefs. Neuroimage 2016.

FitzGerald* T., Schwartenbeck* P. and Dolan R. Reward-related activity in ventral striatum is action-contingent and modulated by behavioural relevance. The Journal of Neuroscience 2014.

* equal contribution