link to homepage

Institute of Neuroscience and Medicine

Navigation and service

Talk by Robert E. Kaas

Department of Statistics, Machine Learning Department and Center for the Neural Basis of Cognition Carnegie Mellon University

25 Sep 2014 13:30
25 Sep 2014 14:30
INM-6, Bldg. 15.22, Seminar Room 3009, 1. OG

Statistical Considerations in Making Inferences about Neural Networks: The Case of Synchrony Detection

Representations of network functional connectivity typically involve nodes (neurons or brain regions) and edges (their co-activation) in a graph. When the number of nodes and edges is large it becomes difficult to make reliable statistical inferences about the details of graphical structure, and the way it changes with stimulus or behavior. I will discuss some strategies used to attack this problem. In particular, I will focus on a measure of neural synchrony that combines point process regression models of individual-neuron activity with loglinear models of multiway synchronous interaction (Kelly and Kass, 2012, Neural Computation), and will describe a method based on Bayesian control of false discoveries that does a good job of distinguishing real from spurious edges in the graph.