Adaptive Multiscale Encoding: A Computational Function of Neuronal Synchronization

Abstract

From the mammalian visual system to high‐dimensional artificial sensor data, the notion of scale is pivotal for information encoding and analysis. This chapter proposes a new, nonlinear mechanism based on neural synchronization that achieves the desired multiscale encoding. Entrainment of different neurons produces larger receptive fields than that of a single cell alone, leading to a multiresolution representation. In vision, the concept of the receptive field is the product of an intrinsically linear approach to visual information processing. However, recent results challenge the notion of linearity, and the idea of a stimulus‐invariant receptive field in general. It has also been suggested that the neural coding is adaptive, that sensory neurons change their responses dynamically, and moreover, that the structure of neuronal correlations changes as well. The neural network model demonstrates a new computational function of synchronization— the multiscale encoding that underlies cognitive and perceptual functions.