Averaging Rate Scheduler Impact on Decentralized Learning with Heterogeneous Data
The author proposes an Averaging Rate Scheduler to address heterogeneity in decentralized learning, demonstrating a 3% improvement in test accuracy compared to conventional methods.
