Manifold transform by recurrent cortical circuit enhances robust encoding of familiar stimuli

TL;DR

The brain becomes more efficient at recognizing familiar faces by adapting neural circuits to focus on important features and ignore irrelevant changes.

Contribution

A computational model showing that recurrent neural circuits develop to robustly encode familiar stimuli through a manifold transform.

Findings

Familiarity training leads to the development of recurrent circuits that robustly encode global stimuli.

The learned circuits implement a manifold transform that enhances stimulus representation against noise.

Local linear computations in the circuit redistribute gain to improve concept discrimination.

Abstract

A ubiquitous phenomenon observed along the ventral stream of the primate hierarchical visual system is the suppression of neural responses to familiar stimuli at the population level. The observation of the suppression of the neural response in the early visual cortex (V1 and V2) to familiar stimuli that are multiple times larger in size than the receptive fields of individual neurons implicates the plausible development of recurrent circuits for encoding these global stimuli. In this work, we investigated the neural mechanisms of familiarity suppression and showed that a recurrent neural circuit based on Hebbian learning, consisting of neurons with small and local receptive fields, can develop to encode specific global familiar stimuli robustly as a result of familiarity training. We proposed that the learned recurrent circuit implements a manifold transform. The recurrent circuit compresses the dimensions of nuisance variations of a familiar image in the neural response manifold relative to the dimensions for discriminating different familiar stimuli, resulting in increased robustness of the global stimulus representation against noise and other irrelevant perturbations. We demonstrate that a recurrent circuit implements the manifold transform using a mixed strategy of locally linear and globally nonlinear computations, where the local linear computation selectively redistributes recurrent gain to enhance concept discrimination. These results provide testable predictions for neurophysiological experiments.

In this research, we explored how the brain can become more efficient at processing familiar visual information. When we repeatedly see something, our brain’s response to it changes. In response to familiar stimuli, neurons across different visual areas of the mammalian visual system become more selective and their overall activities decrease. We developed a computational model to investigate why this happens and what functional advantages these mechanisms might provide. We discovered that familiarity leads to the development of a more efficient and robust neural representation of what we see. It allows us to rapidly and robustly recognize our friend’s face despite changes in lighting conditions, view angle, or facial expression. Our model showed that through repeated exposure, the brain’s neural circuits, even in the early stages of visual processing, rapidly adapt and organize themselves to focus on important and consistent features in our visual environment while becoming less sensitive to irrelevant variations, and distractions.

Figures