Cross-Gradient Aggregation for Decentralized Learning from Non-IID Data

  Yasaman Esfandiari 1 Sin Yong Tan 1 Zhanhong Jiang 2 Aditya Balu 1 Ethan Herron 1 Chinmay Hegde 3
                      Soumik Sarkar 1

             Abstract
   Decentralized learning enables a group of collab-
   orative agents to learn models using a distributed
   dataset without the need for a central parame-
   ter server. Recently, decentralized learning algo-
   rithms have demonstrated state-of-the-art results
   ...