Abstract
This paper presents an analog implementation of probabilistic spiking neural network for portable or biomedical applications which require learning or classification. Online learning adjusts weights by spike based computation. The weight is saved in the long-term synaptic memory. The circuit primarily uses the switched-capacitor structures and was fabricated using 0.18μm CMOS technology. This chip consumes less than 10μW under a 1V supply and the core area of the chip occupies 0.43mm 2 . The chip can learn 80 random patterns with the area under curve of 0.8. The result indicates the chip is appropriate for portable or biomedical applications.