Research map / Optimization / Optimizers
Asynchronous Stochastic Gradient Descent for Extreme-Scale Recommender Systems
Lewis Liu, Kun Zhao
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 |
|---|---|---|---|
| Apply normalized staleness decay to asynchronous gradients | ASGD for the Wide & Deep variant · E-commerce CTR, 3K workers | AUC | Improved |
| Use linear staleness normalization instead of the proposed normalized decay | ASGD for the Wide & Deep variant · E-commerce CTR, 3K workers | AUC | Hurt |
| Add data normalization to staleness normalization | ASGD for the Wide & Deep variant · E-commerce CTR, 1K workers | AUC: improves the result by 1% | Improved |
| Add data normalization to staleness normalization | ASGD for the Wide & Deep variant · E-commerce CTR, 3K workers | AUC: improves the result by 2% | Improved |
| Remove data normalization (ablation) | Staleness-normalized ASGD · E-commerce CTR, 1K and 3K workers | AUC: 1%; 2% | Hurt |
| Use Adagrad-SWAP instead of original Adagrad | Wide & Deep variant · E-commerce CTR, 10 days of incremental training | AUC | Improved |
| Use original Adagrad without SWAP (ablation) | Wide & Deep variant · E-commerce CTR, 10 days of incremental training | AUC | Hurt |
| Decay Adagrad accumulators at every step instead of periodically (ablation) | Wide & Deep variant · E-commerce CTR, 10 days of incremental training | AUC | No clear effect |
| Scale asynchronous training from 1K to 3K workers | ASGD without the proposed staleness normalization · E-commerce CTR | AUC | Hurt |
| Scale asynchronous training from 1K to 3K workers while using staleness normalization | ASGD with staleness normalization · E-commerce CTR | AUC | No clear effect |
Closest papers by meaning
- Addressing Cold-Start Problem in Click-Through Rate Prediction via Supervised Diffusion Modeling (AAAI 2025 · similarity 0.85)
- Ringmaster ASGD: The First Asynchronous SGD with Optimal Time Complexity (ICML 2025 · similarity 0.85)
- Rescaled Asynchronous SGD: Optimal Distributed Optimization under Data and System Heterogeneity (NeurIPS 2026 · similarity 0.85)
- Stability and Generalization of Asynchronous SGD: Sharper Bounds Beyond Lipschitz and Smoothness (NeurIPS 2024 · similarity 0.85)
- On Efficiency-Effectiveness Trade-off of Diffusion-based Recommenders (NeurIPS 2025 · similarity 0.85)
- Fading to Grow: Growing Preference Ratios via Preference Fading Discrete Diffusion for Recommendation (NeurIPS 2025 · similarity 0.84)
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.
