Tackling the Objective Inconsistency Problem in Heterogeneous Federated Optimization
This solves the problem of biased model convergence in federated learning for applications with heterogeneous client data and speeds, though it is incremental as it builds on existing methods like FedAvg and FedProx.
The paper addresses the objective inconsistency problem in heterogeneous federated optimization, where naive aggregation leads to biased solutions, and proposes FedNova, a normalized averaging method that eliminates this inconsistency while maintaining fast convergence.
In federated optimization, heterogeneity in the clients' local datasets and computation speeds results in large variations in the number of local updates performed by each client in each communication round. Naive weighted aggregation of such models causes objective inconsistency, that is, the global model converges to a stationary point of a mismatched objective function which can be arbitrarily different from the true objective. This paper provides a general framework to analyze the convergence of federated heterogeneous optimization algorithms. It subsumes previously proposed methods such as FedAvg and FedProx and provides the first principled understanding of the solution bias and the convergence slowdown due to objective inconsistency. Using insights from this analysis, we propose FedNova, a normalized averaging method that eliminates objective inconsistency while preserving fast error convergence.