DNN-assisted statistical analysis of a model of local cortical circuits

Yaoyu Zhang,Lai-Sang Young

doi:10.1038/s41598-020-76770-3

Abstract

In neuroscience, computational modeling is an effective way to gain insight into cortical mechanisms, yet the construction and analysis of large-scale network models—not to mention the extraction of underlying principles—are themselves challenging tasks, due to the absence of suitable analytical tools and the prohibitive costs of systematic numerical exploration of high-dimensional parameter spaces. In this paper, we propose a data-driven approach assisted by deep neural networks (DNN). The idea is to first discover certain input-output relations, and then to leverage this information and the superior computation speeds of the well-trained DNN to guide parameter searches and to deduce theoretical understanding. To illustrate this novel approach, we used as a test case a medium-size network of integrate-and-fire neurons intended to model local cortical circuits. With the help of an accurate yet extremely efficient DNN surrogate, we revealed the statistics of model responses, providing a detailed picture of model behavior. The information obtained is both general and of a fundamental nature, with direct application to neuroscience. Our results suggest that the methodology proposed can be scaled up to larger and more complex biological networks when used in conjunction with other techniques of biological modeling.

Highlights

In neuroscience, computational modeling is an effective way to gain insight into cortical mechanisms, yet the construction and analysis of large-scale network models—not to mention the extraction of underlying principles—are themselves challenging tasks, due to the absence of suitable analytical tools and the prohibitive costs of systematic numerical exploration of high-dimensional parameter spaces
We present our first key results, which consist of a statistical analysis of the derivatives of model responses and their interpretation
We propose a statistical approach to this problem via the use of deep neural networks (DNN) surrogates

Summary

Introduction

Computational modeling is an effective way to gain insight into cortical mechanisms, yet the construction and analysis of large-scale network models—not to mention the extraction of underlying principles—are themselves challenging tasks, due to the absence of suitable analytical tools and the prohibitive costs of systematic numerical exploration of high-dimensional parameter spaces. The idea is to first discover certain input-output relations, and to leverage this information and the superior computation speeds of the well-trained DNN to guide parameter searches and to deduce theoretical understanding. To illustrate this novel approach, we used as a test case a medium-size network of integrate-and-fire neurons intended to model local cortical circuits. Our proposal is to identify a finite number “inputs” and “outputs” of the model that are important to us—unknown parameters, for example, can be in the “inputs” category—and to first discover, without prejudice, an approximation of the input-output mapping Such a task is well suited to deep neural nets (DNN). Both model building and their statistical analysis are essential steps towards a better understanding of biological mechanisms

Methods

Results

Conclusion