A Predictive Coding Account for Chaotic Itinerancy
This work addresses a theoretical problem in neurorobotics research by linking chaotic itinerancy to predictive coding, but it appears incremental as it applies an existing method (predictive coding) to a new context without broad experimental validation.
The paper tackled the problem of modeling chaotic itinerancy, a phenomenon of autonomous switching between stable behaviors in dynamical systems, by connecting it to predictive coding theory and demonstrating that a recurrent neural network implementing predictive coding can generate neural trajectories similar to chaotic itinerancy under input noise, proposing two scenarios for attractor switching.
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.