On inferring functional connectivity with directed information in neuronal networks

In order to understand modifications in a neural network due to learning, it is paramount to develop an effective tool that is capable of mapping the circuit connectivity and its changes from large-scale recordings of neuronal activity patterns. In this paper, context tree maximizing (CTM) is used to estimate directed information (DI) to measure causal influences among neural spike trains. The method reliably identifies the circuit structures of realistic Hodgkin-Huxley simulated networks. It also produces a promising mapping of a small brain network.