Distributed Optimization: Aggregating Historical Pseudo-Gradients for Improved Convergence
Then, the auxiliary variable 'A' serves as the aggregated historical pseudo-gradient information, which consists of the estimates of the pseudo-gradients 'B' and 'C' from neighboring players within the cluster 'G'.
原文地址: https://www.cveoy.top/t/topic/qA5t 著作权归作者所有。请勿转载和采集!