Training dynamically balanced excitatory-inhibitory networks

Alessandro Ingrosso,L F Abbott,Daniel Bush

doi:10.1371/journal.pone.0220547

Abstract

The construction of biologically plausible models of neural circuits is crucial for understanding the computational properties of the nervous system. Constructing functional networks composed of separate excitatory and inhibitory neurons obeying Dale’s law presents a number of challenges. We show how a target-based approach, when combined with a fast online constrained optimization technique, is capable of building functional models of rate and spiking recurrent neural networks in which excitation and inhibition are balanced. Balanced networks can be trained to produce complicated temporal patterns and to solve input-output tasks while retaining biologically desirable features such as Dale’s law and response variability.

Highlights

Cortical neurons typically require only a small fraction of their thousands of excitatory inputs to reach firing threshold
We have introduced a fast alternative to recursive least squares (RLS) that is capable of training sign-constrained ratebased and spiking network models and, in addition, has the promising features of good memory and computational requirements when dealing with E/I models
We described the conditions under which dynamically balanced networks can be obtained with the training procedure

Summary

Introduction

Cortical neurons typically require only a small fraction of their thousands of excitatory inputs to reach firing threshold This suggests an overabundance of excitation that must be balanced by inhibition to keep neurons within their functional operating ranges. We subdivide tight balance into two classes, parametric and dynamic, depending on whether or not fine tuning of parameters is involved in maintaining the tight balance This is important within the context of our study because, parametrically balanced networks can be constructed and function as models, it is unclear whether the required fine tuning could be accomplished in a biological network. For this reason, we place emphasis on ways of training networks that result in a dynamically balanced configuration

Objectives

Methods

Results

Discussion

Conclusion