Lateral predictive coding revisited: Internal model, symmetry breaking, and response time

TL;DR

This paper revisits lateral predictive coding in neural networks, showing how symmetry breaking affects response times and correlations, with implications for understanding brain information processing.

Contribution

It provides analytical and numerical insights into lateral interactions in predictive coding, emphasizing symmetry breaking and response dynamics in simple neural models.

Findings

Learning breaks interaction symmetry between neurons.

High input correlation does not imply strong direct interactions.

Optimized networks respond faster to familiar inputs and reduce output correlations.

Abstract

Predictive coding is a promising theoretical framework in neuroscience for understanding information transmission and perception. It posits that the brain perceives the external world through internal models and updates these models under the guidance of prediction errors. Previous studies on predictive coding emphasized top-down feedback interactions in hierarchical multi-layered networks but largely ignored lateral recurrent interactions. We perform analytical and numerical investigations in this work on the effects of single-layer lateral interactions. We consider a simple predictive response dynamics and run it on the MNIST dataset of hand-written digits. We find that learning will generally break the interaction symmetry between peer neurons, and that high input correlation between two neurons does not necessarily bring strong direct interactions between them. The optimized network responds to familiar input signals much faster than to novel or random inputs, and it significantly reduces the correlations between the output states of pairs of neurons.