Seminar Title: Identifying internal states from animal decision-making behavior using input-output Hidden Markov Models
Speaker: Jonathan Pillow, PhD.
Professor of Neuroscience, Princeton University
Abstract: Classical models of perceptual decision-making assume that subjects use a single, consistent strategy to form decisions, or that decision-making strategies evolve slowly over time. In this talk, I will present new analyses suggesting that this common view is incorrect. We analyzed data from mouse decision-making experiments and found that decision-making relies on multiple interleaved strategies. These strategies, characterized by states in a hidden Markov model, persist for tens to hundreds of trials before switching, and typically consist of a single "engaged" state, in which mice perform the task with high accuracy, and multiple biased or disengaged states, in which errors are frequent. We also applied this modeling framework to data collected during optogenetic inactivation of the striatum and cortex, which revealed that these circuits make highly state-dependent contributions to decision-making behavior. These results suggest that sensory decision-making relies on multiple state-dependent strategies subserved by distinct neural circuits. Our modeling approach represents a powerful approach to characterizing internal states that we believe will yield new insights into a variety of other behavioral and neural datasets.
About the Speaker: Jonathan completed his undergraduate education at the University of Arizona in Tucson, where he studied mathematics and philosophy. He received a Ph.D. in neuroscience from New York University in 2005, and was postdoctoral fellow at the Gatsby Computational Neuroscience Unit at University College London. In 2009, he became an assistant professor at the University of Texas at Austin, and in 2014 Jonathan moved to Princeton University to join the Princeton Neuroscience Institute, Psychology department, and Center for Statistics & Machine Learning. Jonathan's current research sits at the border between neuroscience and statistical machine learning, and focuses on computational and statistical methods for understanding how large populations of neurons transmit and process information.