Efficient training models of Spiking Neural Networks deployed on a neuromorphic computing architectures

Artificial neural networks (ANN) closest to natural neural networks are a class of Deep Neural Networks (DNN) called Spiking Neural Networks (SNN). They are considered the 3rd (and next) generation of neural networks and have promising behavior for reducing the gap between neuroscience inspired technologies and machine learning technologies, using models of neuron networks for computation. An SNN operates using spikes as time-stamped events for both input and output information. A spike occurrence is modeled by equations that represent a neuron membrane potential, which fires information when it reaches a certain potential (or threshold). The purpose of this paper is to analyze the current solutions for training SNNs, the current solutions for neuromorphic hardware, and to demonstrate there is potentiality in combining them.