Research map / Optimization / Optimizers
Improved Stochastic Optimization of LogSumExp
Egor Gladin, Alexey Kroshnin, Jia-Jie Zhu, Pavel Dvurechensky
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 |
|---|---|---|---|
| replace KL with Safe KL (rho=1e-3) in KL-DRO | KL-DRO SGD · California housing, lambda=1, |D|=10 | objective: 20.0 -> 2.1 | Improved |
| replace KL with Safe KL (rho=1e-3) with large batch | KL-DRO SGD · California housing, lambda=1/5, |D|=1000 | objective: 9.1 -> 25.3 | Hurt |
| use rho=1e-1 instead of 1e-3 (larger rho) (ablation) | Safe KL KL-DRO · California housing, lambda=1, |D|=1000 | objective: 2.5 -> 21.8 | Hurt |
| use rho=1e-5 instead of 1e-3 (smaller rho) (ablation) | Safe KL KL-DRO · California housing, lambda=5, |D|=10 | objective: 0.76 -> 1.03 | Hurt |
| replace KL with Safe KL (rho=1e-1) in UOT-DRO | UOT-DRO SGD · MNIST 25% noise, gamma=1, lambda=1 | objective: 0.81 -> 0.37 | Improved |
| use rho=1 instead of 1e-1 (ablation) | Safe KL UOT-DRO · MNIST 25% noise, gamma=1/5, lambda=1/5 | objective: 1.57 -> 4.76 | Hurt |
| replace KL with Safe KL (rho=1e-1) in UOT-DRO under 85% label noise | UOT-DRO SGD · MNIST 85% noise, gamma=1, lambda=1 | objective: 1.26 -> 0.83 | Improved |
| replace KL with Safe KL under 85% label noise | UOT-DRO SGD · MNIST 85% noise | cross-entropy on hardest class: 2.15 -> 1.87 | Improved |
| replace KL with Safe KL (rho=1e-1) in KL-DRO under distribution shift | KL-DRO SGD · ACS PUMS CA to MS | RMSE on hardest group (x10^3): 75.8 -> 74.7 | Improved |
| replace KL with Safe KL (rho=1e-1) in KL-DRO under distribution shift | KL-DRO SGD · ACS PUMS CA to MS | time per epoch (s): 92 -> 99 | Hurt |
Closest papers by meaning
- Solving Discrete (Semi) Unbalanced Optimal Transport with Equivalent Transformation Mechanism and KKT-Multiplier Regularization (NeurIPS 2025 · similarity 0.86)
- A Geometry-Aware Efficient Algorithm for Compositional Entropic Risk Minimization (ICML 2026 · similarity 0.86)
- Stochastic Optimization for Large-scale Optimal Transport (NeurIPS 2016 · similarity 0.85)
- Efficient Algorithms for Robust and Partial Semi-Discrete Optimal Transport (NeurIPS 2025 · similarity 0.85)
- Stochastic Optimization in Semi-Discrete Optimal Transport: Convergence Analysis and Minimax Rate (NeurIPS 2025 · similarity 0.85)
- Bisimulation Metrics are Optimal Transport Distances, and Can be Computed Efficiently (NeurIPS 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.
