A Spiking Neural Network System for Robust Sequence Recognition

This chapter presents a biologically plausible network architecture with spiking neurons for sequence recognition. This architecture is a unified and consistent system with functional parts of sensory encoding, learning and decoding. This system is the first attempt that helps to reveal the systematic neural mechanisms considering both the upstream and the downstream neurons together. The whole system is consistently combined in a temporal framework, where the precise timing of spikes is considered for information processing and cognitive computing. Experimental results show that our system can properly perform the sequence recognition task with the integration of all three functional parts. The recognition scheme is robust to noisy sensory inputs and it is also invariant to changes in the intervals between input stimuli within a certain range. The classification ability of the temporal learning rule used in our system is investigated through two benchmark tasks including an XOR task and an optical character recognition (OCR) task. Our temporal learning rule outperforms other two benchmark rules that are widely used for classification. Our results also demonstrate the computational power of spiking neurons over perceptrons for processing spatiotemporal patterns.