SC20 Is Everywhere We Are

SC20 Virtual Platform
Accelerate Distributed Stochastic Gradient Descent for Nonconvex Optimization with Momentum
Author/Presenters
Event Type
Workshop
Registration Categories
W
TimeThursday, 12 November 20203:30pm - 4pm EST
LocationTrack 3
DescriptionMomentum method has been used extensively in optimizers for deep learning. Recent studies show that distributed training through K-step averaging has many nice properties. We propose a momentum method for such model averaging approaches. At each individual learner level traditional stochastic gradient is applied. At the meta-level (global learner level), one momentum term is applied and we call it block momentum. We analyze the convergence and scaling properties of such momentum methods. Our experimental results show that block momentum not only accelerates training, but also achieves better results.
Back To Top Button