Modeling of Learning-Induced Receptive Field Plasticity in Auditory Neocortex

Abstract : A motivating assumption of this research is that insights into the structure and operation of the brain can inspire important ideas for the design of intelligent machines, in particular machines that detect and classify auditory and visual signals buried in noise. In this project we focus on the ability of the auditory cortex to adapt its sensitivity to pure tones. Our objective is a model of learning that will enable us to predict changes in the receptive fields of pyramidal cells in the auditory cortex in response to conditioning. Our model is constructed in two major parts: a local process that accounts for the preconditioned states of the pyramidal neurons, and a global process that accounts for the dispersed impact of conditioning across the pyramidal neurons. Recent findings in awake behaving animals have shown that neuronal tuning to acoustic features, e.g., frequency, is systematically altered in the auditory cortex as a result of learning. Responses to training signals are increased whereas responses to other stimuli are decreased, often enough to make the training signal become the most potent stimulus for a cell. This adaptive filtering appears to be a fundamental property of auditory signal processing.