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

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.

Estimating receptive fields of simple and complex cells in early visual cortex: A convolutional neural network model with parameterized rectification

“…Unlike our model, this method estimates different numbers of LN subunits depending on the simple or complex-like nature of the modelled neuron. More recent approaches [18, 25–28] employed deep neural network models, which can be quite complex and challenging to interpret. In all these approaches, the model parameters are not straightforwardly related to the simple/complex distinction.…”

“…These models typically use a succession of layers, each having multiple spatially homogeneous convolutional filters, separated by half-wave rectifiers. Such multi-layer "deep" neural networks (DNNs) have been used to estimate responses of neurons in striate cortex (Cadena et al, 2017;Ecker et al, 2018;Kindel et al, 2019;Zhang & Young, 2020). However, it can be unclear how to interpret intermediate stages of such complicated model architectures.…”

Estimating receptive fields of simple and complex cells in early visual cortex: A convolutional neural network model with parameterized rectification

“…However both the latter approaches would be at the expense of many additional model parameters. More recent approaches [19,[25][26][27][28] employed deep neural network models, which can be quite complex and challenging to interpret. In all these approaches, the model parameters are not straightforwardly related to the simple/complex distinction.…”

“…These models typically use a succession of layers, each having multiple spatially homogeneous convolutional filters, separated by half-wave rectifiers. Such multi-layer "deep" neural networks (DNNs) have been used to estimate responses of neurons in striate cortex [19,[25][26][27][28]. However, it can be unclear how to interpret intermediate stages of such complicated model architectures.…”