Research map / Federated learning and privacy / Communication-efficient FL
Cross-Gradient Aggregation for Decentralized Learning from Non-IID Data
Yasaman Esfandiari, Sin Yong Tan, Zhanhong Jiang, Aditya Balu, Ethan Herron, Chinmay Hegde, Soumik Sarkar
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use cross-gradient aggregation with QP projection instead of standard decentralized aggregation | CIFAR-10 decentralized learning · CIFAR-10 non-IID, fully connected, 5 agents | testing accuracy: 67.3% -> 68.4% | Improved |
| Use cross-gradient aggregation with QP projection instead of standard decentralized aggregation | CIFAR-10 decentralized learning · CIFAR-10 non-IID, fully connected, 10 agents | testing accuracy: 59.3% -> 68.2% | Improved |
| Use cross-gradient aggregation with QP projection instead of standard decentralized aggregation | CIFAR-10 decentralized learning · CIFAR-10 non-IID, fully connected, 40 agents | testing accuracy: 46.2% -> 62.1% | Improved |
| Use cross-gradient aggregation with QP projection instead of standard decentralized aggregation | CIFAR-10 decentralized learning · CIFAR-10 non-IID, ring, 5 agents | testing accuracy: 62.5% -> 66.5% | Improved |
| Use cross-gradient aggregation with QP projection instead of standard decentralized aggregation | CIFAR-10 decentralized learning · CIFAR-10 non-IID, ring, 10 agents | testing accuracy: 38.5% -> 48.8% | Improved |
| Use cross-gradient aggregation with QP projection instead of standard decentralized aggregation | CIFAR-10 decentralized learning · CIFAR-10 non-IID, ring, 40 agents | testing accuracy: 31.4% -> 40.9% | Improved |
| Use cross-gradient aggregation with QP projection instead of standard decentralized aggregation | CIFAR-10 decentralized learning · CIFAR-10 non-IID, bipartite, 5 agents | testing accuracy: 63.9% -> 67.2% | Improved |
| Use cross-gradient aggregation with QP projection instead of standard decentralized aggregation | CIFAR-10 decentralized learning · CIFAR-10 non-IID, bipartite, 10 agents | testing accuracy: 33.8% -> 38.9% | Improved |
| Use cross-gradient aggregation with QP projection instead of standard decentralized aggregation | CIFAR-10 decentralized learning · CIFAR-10 non-IID, bipartite, 40 agents | testing accuracy: 23.5% -> 25.7% | Improved |
| Compress self- and cross-gradients using error-feedback signed gradients | CGA · CIFAR-10 non-IID, fully connected, 5 agents | testing accuracy: 68.4% -> 61.7% | Hurt |
| Compress self- and cross-gradients using error-feedback signed gradients | CGA · CIFAR-10 non-IID, ring, 5 agents | testing accuracy: 66.5% -> 50.3% | Hurt |
| Compress self- and cross-gradients using error-feedback signed gradients | CGA · CIFAR-10 non-IID, bipartite, 5 agents | testing accuracy: 67.2% -> 40.4% | Hurt |
Closest papers by meaning
- Optimal Statistical Rates for Decentralised Non-Parametric Regression with Linear Speed-Up (NeurIPS 2019 · similarity 0.87)
- Decentralized Learning for Overparameterized Problems: A Multi-Agent Kernel Approximation Approach (ICLR 2022 · similarity 0.87)
- DePRL: Achieving Linear Convergence Speedup in Personalized Decentralized Learning with Shared Representations (AAAI 2024 · similarity 0.86)
- Robust Distributed Gradient Aggregation Using Projections onto Gradient Manifolds (AAAI 2024 · similarity 0.86)
- Decentralized Sporadic Federated Learning: A Unified Algorithmic Framework with Convergence Guarantees (ICLR 2025 · similarity 0.85)
- Byzantine-Robust Decentralized Learning via Remove-then-Clip Aggregation (AAAI 2024 · 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.
