Models of Predictive Processing in the Primary Visual Cortex

Abstract

A recent experiment found that neurons in layer 2/3 of the mouse primary visual cortex during the viewing of periodic image sequences exhibited excess activity in response to the presentation of a novel image or sequence, and that this excess activity adapted over the presentation of a sequence. In this work, we construct a neural network model and a probabilistic model that can replicate the activity patterns of these neurons in response to similar stimuli. In doing so, we test a set of hypotheses regarding the mechanisms underlying the activity patterns observed in the aforementioned experimental work, their computational function in the service of inference, and their significance in relation to predictive processing theories of perception.