A neuromorphic categorization system with Online Sequential Extreme Learning

Abstract

This paper presents an event-driven categorization system which processes the address events from a Dynamic Vision Sensor. Using neuromorphic processing, cortex-like spike-based features are extracted by an event-driven MAX-like convolutional network. The extracted spike patterns are then classified by an Online Sequential Extreme Learning Machine with Auto Encoder. Using a Lookup Table, we achieve a virtually fully connected system by physically activating only a very small subset of the classification network. Experimental results show that the proposed system has a very fast training speed while still maintaining a competitive accuracy.