A hardware depressing synapse and its application to contrast-invariant pattern recognition

Analog circuits for depressing synapses are proposed for emulating the dynamic properties of neural networks using dynamic neurons. Although the circuits have few MOS transistors, they mimic well the dynamic properties of depressing synapses. A simple neural network using depressing synapses is introduced for evaluating the performance of hardware depressing synapses. We show that a device using the neural network can perform contrast-invariant pattern recognition based on a neuromorphic processing architecture.