Orientation-selective aVLSI spiking neurons.

摘要: We describe a programmable multi-chip VLSI neuronal system that can be used for exploring spike-based information processing models. The consists of silicon retina, PIC microcontroller, and transceiver chip whose integrate-and-fire neurons are connected in soft winner-take-all architecture. circuit on this multi-neuron approximates cortical microcircuit. configured different computational properties by the virtual connections selected set pixels retina. wiring between chips is effected an event-driven communication protocol uses asynchronous digital pulses, similar to spikes system. synthesize orientation-tuned using both feedforward model feedback model. performance our analog hardware spiking matched experimental observations simulations continuous-valued neurons. has advantages over computer models it real-time, time does not scale with size network.