Massachusetts Institute of Technology
Building and evaluating multi-system functional models of brains
Almost everything we do requires coordination of many brain systems: perception, cognition, action, memory, emotion, etc. Reading this, your visual system processes the text, your language system understands its meaning, and your motor system controls the fingers to scroll down the page. Neuroscientists have built many computer models for individual systems, including perceptual, cognitive, episodic memory and navigation systems, etc. But computer models describing how brain functions involve multiple neural systems are rare. Our group’s research program involves building and evaluating such multi-system integrative models of brain function.
Our goal is to build computer models that are capable of a wide variety of brain functions while integrating and explaining experimental data at several levels. The dominant approach in computational neuroscience is to build models that are focused on individual functions. Our approach complements the dominant paradigm by offering computational models designed to explain neural data over many brain areas across a wide range of brain functions.
Our lab will focus on building neural network models of cognition that learn many tasks efficiently and flexibly. Besides a cognitive system, our models will incorporate several other neural systems (vision, memory, etc.). The cognitive system will learn to integrate information from other systems, and broadcast information back. To evaluate these models, we will set up high-throughput pipelines and develop new benchmarks to compare the behavior, representations, and dynamics of our multi-system networks with large-scale experimental datasets.