Research map / Learning theory and algorithms / Submodular and game algorithms
A Direct Second-Order Method for Solving Two-Player Zero-Sum Games
David Yang, Yuan Gao, Tianyi Lin, Christian Kroer
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 |
|---|---|---|---|
| switch from PRM+ (QA) to semi-smooth Newton (PSSN v2) for high precision | PRM+ (QA) · Random uniform 400x800, gap 1e-10 | clock time (s): 21.97 -> 4.307 | Improved |
| switch to SSN (PSSN v2) at 1e-8 precision | PRM+ (QA) · Random uniform 400x800, gap 1e-8 | clock time (s): 6.424 -> 4.290 | Improved |
| switch to SSN (PSSN v1) at 1e-8 precision | PRM+ (LI) · Random uniform 400x800, gap 1e-8 | clock time (s): 21.69 -> 5.767 | Improved |
| hybrid PSSN v2 at low precision 1e-4 | PRM+ (QA) · Random uniform 400x800, gap 1e-4 | clock time (s): 0.030 -> 0.385 | Hurt |
| hybrid PSSN v2 at 1e-6 precision | PRM+ (QA) · Random uniform 400x800, gap 1e-6 | clock time (s): 0.293 -> 3.638 | Hurt |
| switch to SSN (PSSN v2) at 1e-10 precision | PRM+ (QA) · Random normal 400x800, gap 1e-10 | clock time (s): 19.83 -> 5.077 | Improved |
| hybrid PSSN v1 at 1e-8 precision | PRM+ (QA) · Random normal 400x800, gap 1e-8 | clock time (s): 4.923 -> 5.267 | Hurt |
| tuned damping parameter from PRM+ steps (v2) instead of fixed damping (v1) (ablation) | PSSN · Random uniform 400x800, gap 1e-12 | clock time (s): 5.940 -> 4.323 | Improved |
| tuned damping parameter (v2) instead of fixed damping (v1) (ablation) | PSSN · Random normal 400x800, gap 1e-12 | clock time (s): 5.306 -> 5.094 | No clear effect |
| alternate back and forth between PRM+ and SSN (HPSSN) instead of one-time switch (ablation) | PSSN v2 · Random uniform 400x800, gap 1e-12 | clock time (s): 4.323 -> 5.092 | Hurt |
| alternating PRM+/SSN (HPSSN) instead of one-time switch (ablation) | PSSN v2 · Random normal 400x800, gap 1e-8 | clock time (s): 5.055 -> 4.786 | Mixed |
Closest papers by meaning
- Towards convergence to Nash equilibria in two-team zero-sum games (ICLR 2023 · similarity 0.90)
- Competitive Gradient Descent (NeurIPS 2019 · similarity 0.90)
- \(\varepsilon\)-Optimally Solving Two-Player Zero-Sum POSGs (NeurIPS 2025 · similarity 0.89)
- Fast Last-Iterate Convergence of Learning in Games Requires Forgetful Algorithms (NeurIPS 2024 · similarity 0.89)
- Last-Iterate Convergence Properties of Regret-Matching Algorithms in Games (ICLR 2025 · similarity 0.88)
- Solving Zero-Sum Convex Markov Games (ICML 2025 · similarity 0.88)
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.
