Identification of Neural Functional Connectivity Using a Sparse Generalized Volterra Model

• Activity Number: 360
• Type: Contributed
• Date/Time: Tuesday, August 3, 2010 : 10:30 AM to 12:20 PM
• Sponsor: Biometrics Section
• Author(s): Dong Song*+ and Theodore W. Berger
• Companies: University of Southern California and University of Southern California
• Keywords: spike ; GLM ; L1 regularization ; Volterra kernel

Abstract

To understand how a brain region processes information, it is essential to identify the functional connectivity between populations of neurons from their spiking activity. We formulate a physiologically-plausible neuron model that consists of a set of feedforward Volterra kernels, a feedback Volterra kernel, a Gaussian noise, and a threshold. This model is equivalent to a cascade of a multiple-input Volterra model and a probit link function and thus can be termed as a generalized Volterra model (GVM). GVMs are estimated with a group L1-regularization method that results in sparse GVMs. A sparse GVM includes only the significant groups of coefficients corresponding to the significant input-output functional connections and their orders of nonlinearities, and thus provides a quantitative representation of the neural functional connectivity.