Dr. Michael W. Cole, Associate Professor, Center for Molecular & Behavioral Neuroscience, Rutgers University-Newark, https://www.colelab.org
Title: Brain Network Flows and the Generation of Cognitive Abstractions
Abstract: From navigating a new city to learning new tasks, human brains must utilize a wide variety of representations to accomplish their goals. Understanding the origin and coordination of these representations will require determining how such task-relevant information is generated by brain network interactions. I will share recent insights into these processes gained via activity flow modeling – an approach for creating neural network simulations uniquely tied to empirical brain data. This involves using empirical brain connectivity data to build computational models that generate task-evoked brain activity. Investigating the resulting model yields insights into the processes in the brain that generate that activity. The key construct bridging data and modeling is activity flow – the movement of neural activity over brain connections – which is key to computation in neural network models and is known to occur in real brains. I will share results that use the (open source) Brain Activity Flow Toolbox (https://colelab.github.io/ActflowToolbox/) to identify brain network processes underlying the generation of visual categories (e.g., faces; see Figure), highly abstract cognition (e.g., rules), and motor responses (behavior). This approach is also applied to understand the brain network processes underlying aberrant cognition in brain disorders such as schizophrenia and Alzheimer’s disease. Together these findings demonstrate the utility of the activity flow mapping framework for discovering the network processes generating (healthy and unhealthy) cognition in the human brain, ultimately supporting the hypothesis that neural functions are generated (computed) by distributed activity flow processes that are specified primarily by connectivity patterns.
