Aggregation on Learnable Manifolds for Asynchronous Federated Optimization
This addresses performance and fairness problems in asynchronous federated learning for applications like healthcare and general-purpose ML, though it appears incremental as a hybrid of existing geometric and federated learning ideas.
The paper tackles two key issues in asynchronous federated learning with heterogeneous clients: loss barriers from linear parameter interpolation and interference from stale updates. The proposed geometric framework with AsyncBezier and OrthoDC components consistently improves accuracy and client fairness on three datasets including LEAF Shakespeare and FEMNIST, preserving gains even when baselines get more compute.
Asynchronous federated learning (FL) with heterogeneous clients faces two key issues: curvature-induced loss barriers encountered by standard linear parameter interpolation techniques (e.g. FedAvg) and interference from stale updates misaligned with the server's current optimisation state. To alleviate these issues, we introduce a geometric framework that casts aggregation as curve learning in a Riemannian model space and decouples trajectory selection from update conflict resolution. Within this, we propose AsyncBezier, which replaces linear aggregation with low-degree polynomial (Bezier) trajectories to bypass loss barriers, and OrthoDC, which projects delayed updates via inner product-based orthogonality to reduce interference. We establish framework-level convergence guarantees covering each variant given simple assumptions on their components. On three datasets spanning general-purpose and healthcare domains, including LEAF Shakespeare and FEMNIST, our approach consistently improves accuracy and client fairness over strong asynchronous baselines; finally, we show that these gains are preserved even when other methods are allocated a higher local compute budget.