A Predictive Coding Account for Chaotic Itinerancy

TL;DR

This paper demonstrates how a recurrent neural network based on predictive coding can replicate chaotic itinerancy, a phenomenon of autonomous switching between stable states, by generating neural trajectories with input noise.

Contribution

It introduces a novel connection between chaotic itinerancy and predictive coding, showing how neural networks can produce autonomous attractor switching.

Findings

Recurrent predictive coding networks can generate chaotic itinerant trajectories.

Input noise induces autonomous switching between stable states.

Two scenarios for attractor switching are proposed.

Abstract

As a phenomenon in dynamical systems allowing autonomous switching between stable behaviors, chaotic itinerancy has gained interest in neurorobotics research. In this study, we draw a connection between this phenomenon and the predictive coding theory by showing how a recurrent neural network implementing predictive coding can generate neural trajectories similar to chaotic itinerancy in the presence of input noise. We propose two scenarios generating random and past-independent attractor switching trajectories using our model.