Abstract:Neurons in the primary visual cortex respond selectively to simple features of visual stimuli, such as orientation and spatial frequency. Simple cells, which have phase-sensitive responses, can be modeled by a single receptive field filter in a linear-nonlinear model. However, it is challenging to analyze phase-invariant complex cells, which require more elaborate models having a combination of nonlinear subunits. Estimating parameters of these models is made additionally more difficult by cortical neurons' trial-to-trial response variability. We develop a simple convolutional neural network method to estimate receptive field models for both simple and complex visual cortex cells from their responses to natural images. The model consists of a spatiotemporal filter, a parameterized rectifier unit (PReLU), and a two-dimensional Gaussian "map" of the receptive field envelope. A single model parameter determines the simple vs. complex nature of the receptive field, capturing complex cell responses as a summation of homogeneous subunits, and collapsing to a linear-nonlinear model for simple type cells. The convolutional method predicts simple and complex cell responses to natural image stimuli as well as grating tuning curves. The fitted models yield a continuum of values for the PReLU parameter across the sampled neurons, showing that the simple/complex nature of cells can vary in a continuous manner. We demonstrate that complex-like cells respond less reliably than simple-like cells. However, compensation for this unreliability with noise ceiling analysis reveals predictive performance for complex cells proportionately closer to that for simple cells. Most spatial receptive field structures are well fit by Gabor functions, whose parameters confirm well-known properties of cat A17/18 receptive fields. Methods for recording increasingly many visual cortex neurons are advancing rapidly, demanding new approaches to characterize diverse receptive fields. We present a compact convolutional neural network model of early cortical neurons, which is uniformly applicable to simple and complex cells, and whose parameters are straightforwardly interpretable and readily estimated from responses to natural image stimuli. This novel approach introduces a single estimated model parameter to capture the simple/complex nature of a neuron's receptive field, revealing a continuum of simple vs complex-like behaviour. We show that almost all complex-like cells exhibit a lower response reliability to repeated presentations of the same stimuli, compared to more reliable responses from simple-like cells. Accounting for this "noise ceiling" brings predictive performance for complex cells proportionately closer to that for simple cells. Using this model estimation approach with natural images, we evaluate findings from previous approaches that were restricted to simple-type cells and the use of grating or white noise stimuli, revealing a diversity of Gabor-like spatial receptive field shapes, which lie along a continuum of spatial bandwidths.

Development of Invariant Feature Maps Via a Computational Model of Simple and Complex Cells.

Learning receptive field properties of complex cells in V1

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

An Egalitarian Network Model for the Emergence of Simple and Complex Cells in Visual Cortex

Emergence of Complex-Like Cells in a Temporal Product Network with Local Receptive Fields

A New Complex Cell-Inspired Feature Extraction Method

Computational Modeling of the Dynamics of Simple and Complex Cells in Primary Visual Cortex

Bipartite invariance in mouse primary visual cortex

Pooling strategies in V1 can account for the functional and structural diversity across species

Feature Extraction Inspired by V1 in Visual Cortex

Simple and complex cells revisited: toward a selectivity-invariance model of object recognition

Learning Slowness in a Sparse Model of Invariant Feature Detection

Toward a Biologically Plausible Model of LGN-V1 Pathways Based on Efficient Coding

Emergence of simple-cell receptive field properties by learning a sparse code for natural images

Learning Invariant Responses to the Natural Transformations of Objects

Learning V1 Simple Cells with Vector Representation of Local Content and Matrix Representation of Local Motion

The Representation of Complex Images in Spatial Frequency Domains of Primary Visual Cortex.

An Image Representation of Infrastructure Based on Non-Classical Receptive Field

Pooling in a predictive model of V1 explains functional and structural diversity across species

Invariant recognition of feature combinations in the visual system

Efficient Visual Coding: From Retina To V2