A Unified information flow model of Poisson-type brain neuronal network activity

Understanding information transfer and representation in the brain is one of the most challenging scientific endeavors since neuroscientists are still far from converging to a solution with exact description. The main challenge in deciphering the neural code is the probabilistic nature of the neural codebook which maps stimuli to neural responses and vice versa. With the advent of recording techniques from single and multiple cells, it is becoming plausible to obtain large amount of experimental data against which theoretical models can be tested. In this respect, information theory can potentially provide a powerful framework for analyzing information content and representation in neuronal network activity. In this paper, we develop a simplified neuronal firing model which is mathematically complete and scalable. The model makes use of Poisson processes¿ properties and queuing theory. We also derive theoretical tools to measure, quantify, and set upper limits on information coded by the proposed model based on information theory. The designed measures take into account a general coding scheme in each neuron that combines temporal and rate coding of spike train responses of neurons. The proposed model along with the tools are generalized to quantify information in a population of neurons where correlation among neurons is modeled. The model also incorporates a spatial component which allows studying the amount of information gained from the spatial pattern of responses in a neuron population. Accordingly, the developed model and tools aim at providing a unified view of measuring information quantities and hence giving a better understanding of the neural coding.