NeurIPS 2019
Sun Dec 8th through Sat the 14th, 2019 at Vancouver Convention Center
Paper ID:4549
Title:Robust and Communication-Efficient Collaborative Learning

This paper considers distributed solving of collaborative learning problem in the presence of stragglers and communication cost. The two main ideas are quantization of gradients and time limit on local gradient computation. Overall, this paper considers an important problem and solves two main issues in its implementation.