Building a Dynamical Network Model from Neural Spiking Data: Application of Poisson Likelihood

TL;DR

This paper presents a theoretical approach to modeling neural networks using point process likelihood methods, addressing challenges in neural data measurement and providing simulation-supported insights.

Contribution

It introduces a novel application of Poisson likelihood methods for dynamical neural network modeling from spike data, with comparative analysis.

Findings

Effective modeling of neural responses using point process likelihoods

Simulation results validate the proposed modeling approach

Comparison with existing methods highlights advantages

Abstract

Research showed that, the information transmitted in biological neurons is encoded in the instants of successive action potentials or their firing rate. In addition to that, in-vivo operation of the neuron makes measurement difficult and thus continuous data collection is restricted. Due to those reasons, classical mean square estimation techniques that are frequently used in neural network training is very difficult to apply. In such situations, point processes and related likelihood methods may be beneficial. In this study, we will present how one can apply certain methods to use the stimulus-response data obtained from a neural process in the mathematical modeling of a neuron. The study is theoretical in nature and it will be supported by simulations. In addition it will be compared to a similar study performed on the same network model.