Gradient Decomposition Methods for Training Neural Networks With Non-ideal Synaptic Devices
Sign Up to like & getrecommendations! Published in 2021 at "Frontiers in Neuroscience"
DOI: 10.3389/fnins.2021.749811
Abstract: While promising for high-capacity machine learning accelerators, memristor devices have non-idealities that prevent software-equivalent accuracies when used for online training. This work uses a combination of Mini-Batch Gradient Descent (MBGD) to average gradients, stochastic rounding… read more here.
Keywords: streaming batch; rank seq; memristor; decomposition methods ... See more keywords