Research map / Optimization / Optimizers
GAL: Gradient Assisted Learning for Decentralized Multi-Organization Collaborations
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 GAL instead of Alone | local model · CIFAR10 (M=8) | accuracy: 46.3 -> 74.3 | Helped |
| Use GAL instead of Alone | local model · MNIST (M=8) | accuracy: 24.2 -> 96.3 | Helped |
| Use GAL instead of AL (sequential assisted learning) | AL · CIFAR10 (M=8) | accuracy: 51.1 -> 74.3 | Helped |
| Use Deep Model Sharing (shared feature extractor) (ablation) | GAL · CIFAR10 (M=8) | accuracy: 74.3 -> 67.0 | Hurt |
| Use Deep Model Sharing (ablation) | GAL · MNIST (M=8) | accuracy: 96.3 -> 96.3 | No clear effect |
| Use Deep Model Sharing (ablation) | GAL · ShapeNet55 | accuracy: 84.1 -> 85.3 | Helped |
| Use Deep Model Sharing (ablation) | GAL · MIMICL | MAD: 91.9 -> 97.7 | Hurt |
| Use gradient assistance weights under noise sigma=1 (ablation) | GAL · CIFAR10 | accuracy: 45.4 -> 61.0 | Helped |
| Use gradient assistance weights under noise sigma=5 (ablation) | GAL · CIFAR10 | accuracy: 23.3 -> 57.3 | Helped |
| Use gradient assistance weights under noise sigma=5 (ablation) | GAL · MNIST | accuracy: 33.8 -> 92.1 | Helped |
| Local loss l_1 instead of l_2 (ablation) | GAL · CIFAR10 (M=8) | accuracy: 74.3 -> 27.8 | Hurt |
| Local loss l_4 instead of l_2 (ablation) | GAL · MNIST (M=8) | accuracy: 96.3 -> 98.1 | Helped |
Closest papers by meaning
- DePRL: Achieving Linear Convergence Speedup in Personalized Decentralized Learning with Shared Representations (AAAI 2024 · similarity 0.85)
- Free-Rider and Conflict Aware Collaboration Formation for Cross-Silo Federated Learning (NeurIPS 2024 · similarity 0.84)
- Widening the Network Mitigates the Impact of Data Heterogeneity on FedAvg (ICML 2025 · similarity 0.84)
- FedQS: Optimizing Gradient and Model Aggregation for Semi-Asynchronous Federated Learning (NeurIPS 2025 · similarity 0.83)
- Learning to Specialize: Joint Gating-Expert Training for Adaptive MoEs in Decentralized Settings (NeurIPS 2025 · similarity 0.83)
- Unveiling the Power of Multiple Gossip Steps: A Stability-Based Generalization Analysis in Decentralized Training (NeurIPS 2025 · similarity 0.83)
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.