The Bayesian Brain: Theory, Technology and Application

March 2, 2015
Speaker: Rosalyn Moran, Assistant Professor, Bradley Department of Electrical & Computer Engineering, College of Engineering, Virginia Tech


In this talk I will present Bayesian perspectives on the human brain, both as a methodology to assess brain connectivity and as a theoretical framework that formally links synaptic function to cognition and behavior.

In the first part of my talk, I will introduce Dynamic Causal Modeling (DCM) as a ‘mathematical microscope’ for assessing functional brain networks. Using noninvasive neuroimaging data, I will demonstrate how biologically motivated generative models can be deployed with approximate (variational) Bayesian inference techniques to infer upon the complex and latent neuronal architectures that subtend observed time-series data from fMRI, and electrophysiology. Using examples from pathological and pharmacologically-altered cortical circuits, I will show how DCM can also help elucidate the key parameters that contribute to abnormal brain function.

In the second part of my talk I will present a mathematical deconstruction of age-related changes in cortical processing motivated by the Free Energy Principle. This principle hypothesizes a simple optimization that the brain may perform and a potential implementation based on predictive coding. From this perspective, the brain itself represents a model of its environment and offers predictions about the world through a subset of cortical connections, while responding - through learning - to novel interactions and experiences. I will provide evidence for selective alterations in these predictive and updating processes over the lifespan and examine potential adaptive and maladaptive consequences in neurodegenerative disease. Overall, the talk will cover how the brain could ‘do inference’ on the environment, and how scientists can ‘do inference’ on the brain.

Speaker Bio

Rosalyn MoranRosalyn Moran is an assistant professor at Virginia Tech Carilion Research Institute, at the Bradley Department of Electrical & Computer Engineering and at the department of Psychiatry and Behavioral Medicine at Virginia Tech Carilion School of Medicine. Her research employs both theoretical and empirical neuroscientific approaches to understand the principles of functional integration in the brain. In particular, her lab investigates how different neurotransmitters & neuromodulatory systems shape the dynamics of neuronal communication in human brains. Her work involves the development of Bayesian approaches to neuroimaging data and behavioral analysis. She is a co-author of the neuroimaging freeware package Statistical Parametric Mapping (SPM) and serves on its faculty at courses internationally.

Hosted by Aurel A. Lazar.

500 W. 120th St., Mudd 1310, New York, NY 10027    212-854-3105               
©2014 Columbia University