A Deep Unsupervised Feature Learning Spiking Neural Network with Binarized Classification Layers for EMNIST Classification using SpykeFlow

摘要: End user AI is trained on large server farms with data collected from the users. With ever increasing demand for IOT devices, there a need deep learning approaches that can be implemented (at edge) in an energy efficient manner. In this work we approach using spiking neural networks. The unsupervised technique of spike timing dependent plasticity (STDP) binary activations are used to extract features input data. Gradient descent (backpropagation) only output layer perform training classification. accuracies obtained balanced EMNIST set compare favorably other approaches. effect stochastic gradient (SGD) approximations capabilities our network also explored.