Nature versus Nurture: EVOLUTIONARY AND DEVELOPMENTAL ORIGINS OF HYPERPARAMETERS IN INFANT AGENTS

Abstract

The increasing evidence for an organisation of the brain based on the principles of Bayesian inference has led to a rise in popularity of predictive processing (Kwisthout, Bekkering, & van Rooij, 2017). Most research within the predictive processing framework focuses on explaining cognitive phenomena within the adult brain, missing to explain how these phenomena arise and develop in infant agents. We propose a shift of focus in the predictive processing literature towards development. Before we can investigate development, the initial state of generative models needs to be understood. This research attempts to make a starting step by investigating developmental and evolutionary origins of hyperparameters within infant agents.