Improving efficiency of a Stochastic Computing-based Morphological Neural Network

In this work we present an enhancement of a neural network hardware implementation based on an efficient combination of Stochastic Computing (SC) and Morphological Neural Networks (MNN). The enhancement has concentrated on extending an original tiny two-layer network to the more de-manding MNIST benchmark and also pruning up to a 92% of the weights of the morphological layer, allowing a drastic shrinkage of the hardware resources and power dissipation without barely degrading test accuracy. This new proposal contributes to foster a promising ultra-low power Machine Learning methodology such as SC-based MNN are.