Research map / Optimization / Distributed and compressed
DPA: Decentralized Primal Averaging with Quasi-Global Momentum for Highly Heterogeneous Data
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's abstract only. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Proposes Decentralized Primal Averaging (DPA), a decentralized training method extending centralized primal averaging with quasi-global momentum to handle highly heterogeneous client data on sparse communication graphs.
Problem
Decentralized training of deep models across clients with heterogeneous local data degrades test accuracy, especially on sparse communication graphs.
Approach
Extends centralized primal averaging to the decentralized setting and combines it with quasi-global momentum to counter data heterogeneity over sparse graphs.
Experimental setup
Metrics: test accuracy
Builds on: primal averaging quasi-global momentum
Closest papers by meaning
- Quasi-global Momentum: Accelerating Decentralized Deep Learning on Heterogeneous Data (ICML 2021 · similarity 0.91)
- $\textbf{A}^2\textbf{CiD}^2$: Accelerating Asynchronous Communication in Decentralized Deep Learning (NeurIPS 2023 · similarity 0.88)
- Decentralized Sporadic Federated Learning: A Unified Algorithmic Framework with Convergence Guarantees (ICLR 2025 · similarity 0.86)
- DIMAT: Decentralized Iterative Merging-And-Training for Deep Learning Models (CVPR 2024 · similarity 0.85)
- Unveiling the Power of Multiple Gossip Steps: A Stability-Based Generalization Analysis in Decentralized Training (NeurIPS 2025 · similarity 0.85)
- From Promise to Practice: Realizing High-performance Decentralized Training (ICLR 2025 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.