Peter Dayan
Session: D. Computation and cognition: from physiology to neural processing
Will talk about: Neurocomputational Modeling in Psychiatry
Peter Dayan is the director of the Gatsby Computational Neuroscience Unit at University College London. He is co-author of "Theoretical Neuroscience", a leading textbook in computational and mathematical modeling of brain function (see Computational Neuroscience). He is known for applying Bayesian methods from Machine Learning and Artificial Intelligence to understand neural function, and is particularly renowned for having related neurotransmitter levels to reinforcement learning prediction errors in and Bayesian uncertainties. He began his career studying Mathematics at the University of Cambridge (UK) and then continued for a PhD in Cognitive Science at the University of Edinburgh with David Willshaw, which focused on associative memory and reinforcement learning. He then went on to postdoc positions first with Terry Sejnowski at the Salk Institute and then Geoff Hinton at the University of Toronto. He took up an Assistant Professor position at the Massachusetts Institute of Technology, and later moved to University College London to help found the Gatsby Computational Neuroscience Unit.
The nascent field of computational psychiatry includes the application of computational and statistical methods to understand dysfunction in psychiatric disease through the medium of normal information processing and cognition. I will discuss our attempts to use Bayesian decision theory to provide a framework for understanding disorders - coarsely, solving the wrong problem, solving the right problem incorrectly, or solving the right problem correctly, but in the wrong environment. This will be illustrated using examples drawn partly from the case of depression.
