Silicon spiking neurons for hardware implementation of extreme learning machines

In this paper, we propose a silicon implementation of extreme learning machines (ELM) using spiking neural circuits. The major components of a silicon spiking neural network, neuron, synapse and ‘Address Event Representation’ (AER) for asynchronous spike based communication, are described. The benefits of using this hardware to implement an ELM as opposed to other single layer feedforward networks (SLFN) are explained. Several possible architectures for efficient implementation of ELM using these circuits are presented and their possible impact on ELM performance is discussed.