Spike-Based MAX Networks for Nonlinear Pooling in Hierarchical Vision Processing

Complex cells in the visual cortex utilize a maximum (MAX) operation to pool the outputs of simple cells to achieve feature specificity and invariance. We demonstrate a biologically-plausible MAX network for nonlinear pooling in hardware, using a reconfigurable multichip address event representation based VLSI system. With this implementation we have shown that we can implement simple and advanced stages of visual processing on the same chip and are one step closer to constructing an autonomous, continuous-time, biologically- plausible hierarchical model of visual information processing using large-scale arrays of identical silicon neurons.