Research map / Optimization / Distributed and compressed
Statistically Preconditioned Accelerated Gradient Method for Distributed Optimization
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 acceleration with statistical preconditioning (SPAG) rather than DANE | Distributed ERM · RCV1, λ=10^-7 | communication rounds to convergence | Helped |
| Use acceleration with statistical preconditioning rather than DANE | Distributed ERM · KDD2010, λ=10^-7 | convergence speed | Helped |
| Use SPAG rather than AGD | Distributed ERM · KDD2010 | convergence speed | Helped |
| Use acceleration rather than DANE | Distributed ERM · RCV1, λ=10^-5, n=10^5 | convergence rate | No clear effect |
| Increase the preconditioning sample size (ablation) | Distributed ERM · RCV1, λ=10^-5 | convergence speed | Helped |
| Use SPAG rather than HB-DANE | Distributed ERM · RCV1 | convergence speed | Helped |
| Use acceleration rather than non-accelerated DANE | Distributed ERM · RCV1 and KDD2010 | convergence speed | Mixed |
| Vary the preconditioner regularization μ (ablation) | SPAG · RCV1, λ=10^-7, n=10^4 | convergence speed | Mixed |
| Tune the step size separately for each μ (ablation) | SPAG · RCV1, λ=10^-7, n=10^4 | convergence speed | No clear effect |
| Enable gain line search for G_t (ablation) | SPAG · Experiments on RCV1 and KDD2010 | convergence speed | No clear effect |
| Use finer-tuned momentum (ablation) | SPAG and HB-DANE · RCV1 | convergence speed | Helped |
Closest papers by meaning
- Accelerated Methods with Compressed Communications for Distributed Optimization Problems Under Data Similarity (AAAI 2025 · similarity 0.89)
- Hybrid Decentralized Optimization: Leveraging Both First- and Zeroth-Order Optimizers for Faster Convergence (AAAI 2025 · similarity 0.88)
- Spectral Preconditioning for Gradient Methods on Graded Non-convex Functions (ICML 2024 · similarity 0.87)
- Nonlinearly Preconditioned Gradient Methods: Momentum and Stochastic Analysis (NeurIPS 2025 · similarity 0.87)
- Scalable Distributed Stochastic Optimization via Bidirectional Compression: Beyond Pessimistic Limits (NeurIPS 2026 · similarity 0.87)
- Federated Optimization with Doubly Regularized Drift Correction (ICML 2024 · similarity 0.87)
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.