Multilayer In-Place Learning Networks with Adaptive Lateral Connections: Models and Simulations

In this paper, the Multilayer In-place Learning Networks with Adaptive Lateral Connections (MILN-ALC) is proposed to develop both afferent and lateral connections simultaneously, simulating the development of primary visual cortex through sensory experiment. The experimental results using isotropic Gaussian spots and grating indicate that the lateral connections are statistically strongly correlated among nearby neurons, and also show that both afferent and lateral connections depend on input patterns.