Partial information decomposition reveals that synergistic neural integration is greater downstream of recurrent information flow in organotypic cortical cultures.

Samantha P Sherrill,Nicholas M Timme,Ehren L Newman,John M Beggs,Boris S Gutkin

doi:10.1371/journal.pcbi.1009196

Abstract

The directionality of network information flow dictates how networks process information. A central component of information processing in both biological and artificial neural networks is their ability to perform synergistic integration–a type of computation. We established previously that synergistic integration varies directly with the strength of feedforward information flow. However, the relationships between both recurrent and feedback information flow and synergistic integration remain unknown. To address this, we analyzed the spiking activity of hundreds of neurons in organotypic cultures of mouse cortex. We asked how empirically observed synergistic integration–determined from partial information decomposition–varied with local functional network structure that was categorized into motifs with varying recurrent and feedback information flow. We found that synergistic integration was elevated in motifs with greater recurrent information flow beyond that expected from the local feedforward information flow. Feedback information flow was interrelated with feedforward information flow and was associated with decreased synergistic integration. Our results indicate that synergistic integration is distinctly influenced by the directionality of local information flow.

Highlights

Feedforward, recurrent and feedback connections are important for information processing in both artificial and biological neural networks [1,2,3,4,5]
We identified effective connections between neurons in each recording as those that had significant transfer entropy such that the observed value was greater than 99.9% of values obtained from a jittering procedure (i.e. p
We quantified the amount of synergistic integration performed by the receiver based on its inputs using ‘synergy,’ a term derived from partial information decomposition

Summary

Introduction

Feedforward, recurrent and feedback connections are important for information processing in both artificial and biological neural networks [1,2,3,4,5]. A component of information processing that is central to both biological and artificial neural networks is their ability to perform synergistic integration, a form of computation. Understanding how each of these directed functional connections influences the computational properties of neural networks is a critical step in understanding how neural networks compute. We examine this in the context of cortical networks, using a motif-style, information theoretic analysis of high-density in vitro recordings of spiking neurons

Methods

Results

Discussion

Conclusion