Maximum Caliber Infers Effective Coupling and Response from Spiking Networks

TL;DR

This paper introduces a novel application of the Maximum Caliber principle to infer synaptic interactions and neural response properties from spiking activity, enabling effective network characterization.

Contribution

It develops a framework using Maximum Caliber for unbiased inference of neural network parameters from spiking data, including synaptic coupling and response functions.

Findings

Successfully inferred synaptic coupling strengths from simulated data

Reconstructed inter-spike interval distributions accurately

Applied method to salamander retina data with promising results

Abstract

The characterization of network and biophysical properties from neural spiking activity is an important goal in neuroscience. A framework that provides unbiased inference on causal synaptic interaction and single neural properties has been missing. Here we applied the stochastic dynamics extension of Maximum Entropy -- the Maximum Caliber Principle -- to infer the transition rates of network states. Effective synaptic coupling strength and neuronal response functions for various network motifs can then be computed. The inferred minimal model also enables leading-order reconstruction of inter-spike interval distribution. Our method is tested with numerical simulated spiking networks and applied to data from salamander retina.