Complex Function without Complex Structure: Constraining Models in the Era of Big Data

Audrey Sederberg, Ph.D.
Department of Physics
Emory University

Join BlueJeans Meeting

ABSTRACT
A central goal of neuroscience is to connect structure to function: to understand how neural activity and neuroanatomy control the actions, perception, and cognition of an organism. In recent years, there has been an explosion in the quantity and quality of neural data. Only ten years ago, recording simultaneously from a few dozen cells was notable, and now that number is in the thousands. We also know more about the intricacies of microcircuit anatomy, with detailed information on individual cell types and the patterns of connectivity among them. These data are exciting, but also raise challenging questions and require integrating precise, quantitative predictions into the analysis of large, complex datasets. In my talk, I will focus on two examples of how new directions in theoretical and data-analytic research can lead to novel insight into the function of neural circuits. First, I will show how minimally structured networks can capture many features of large-scale neural population recordings with surprising precision (within a few percent!), suggesting new approaches for linking structure to function. In the second part of the talk, I will show how we use prediction to extract essential features of a dynamic cortical state, a general approach that can be extended across brain areas and species to build a quantitative, comparative framework for the analysis of cortical dynamics. These are steps toward the ultimate goal of predicting, from the anatomy of a microcircuit, both the statistics of activity (e.g., selectivity, correlations, power spectra) that it generates and how that activity supports microcircuit computations relevant to behavior.