Learning real-world stimuli by single-spike coding and tempotron rule

In this paper, a system model is built for pattern recognition by using spiking neurons. The system contains encoding, learning and readout. The schemes used in this network are efficient and biologically plausible. Through the encoding of our network, the external stimuli (images) are converted into spatiotemporal spiking patterns. These spiking patterns are then efficiently learned through a supervised temporal learning rule. Through simulation, the properties of the system model are shown. It turns out that this network can successfully recognize different patterns very fast.