A columnar V1/V2 visual cortex model and emulation

Abstract

We have explored the implementation of neurophysiological and psychological constructs to develop a hyper-parallel computing platform. This approach is termed neuromorphic computing. As part of that effort, the primary visual cortex (V1) has been modeled in a high performance computing facility. The current columnarV1 model is being expanded to include binocular disparity and motion perception. Additionally, V2 thick and pale stripes are being added to produce a V1/V2 stereomotion and form perception system. Both the V1 and V2 models are based upon structures approximating neocortical minicolumns and functional columns. The neuromorphic strategies employed include columnar organization, integrate-and-fire neurons, temporal coding, point attraction recurrent networks, Reichardt detectors and “confabulation” networks. The interest is driven by the value of applications which can make use of highly parallel architectures we expect to see surpassing one thousand cores per die in the next few years.A central question we seek to answer is what the architecture of hyper-parallel machines should be. We also seek to understand computational methods akin to how a brain deals with sensation, perception, memory, attention decision-making.