Yuki Ikeda(Supplementary material) Analysis on optimal parameters for existing algorithms and their new…4. Analysis on optimal parameters for existing algorithms and their new combinationsApr 1, 2022Apr 1, 2022
Yuki Ikeda(Supplementary material) Explanations for three local-update methodsEASGD prepares auxiliary variables z, and each step all worker nodes send their current parameters to the auxiliary variables and instead…Apr 1, 2022Apr 1, 2022
Yuki IkedaCommunication-Efficient SGD AlgorithmsThere are many algorithms for distributed training in deep learning, but still there are some open problems such as the following: — a…Mar 31, 2022Mar 31, 2022