Research map / Learning theory and algorithms / Matrix and tensor methods
Zap Q-Learning
Adithya M Devraj, Sean Meyn
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 |
|---|---|---|---|
| use two time-scale matrix gain (rho=0.85) instead of single time-scale (rho=1.0) (ablation) | Zap Q-learning · Finance optimal stopping, outliers h<=0.999 (N=1000 runs) n=2e4 | outlier percentage: 35.7 -> 0.13 | Improved |
| Zap Q-learning matrix gain instead of G-Q(0) matrix gain | G-Q(0) · Finance optimal stopping, outliers h<=0.999 (N=1000 runs) n=2e6 | outlier percentage: 68 -> 0 | Improved |
| Zap Q-learning instead of G-Q(0) | G-Q(0) · Finance optimal stopping, outliers h<=0.999 (N=1000 runs) | asymptotic covariance trace | Improved |
| two time-scale Zap Q-learning vs one time-scale (ablation) | Zap Q-learning · path-finding MDP beta=0.99 | max Bellman error | Improved |
| Zap Q-learning vs Watkins, speedy, polynomial, RPJ | Watkins Q-learning · path-finding MDP beta=0.99 | max Bellman error at n=1e6 | Improved |
| Zap Q-learning vs scalar-gain algorithms | Watkins Q-learning · path-finding MDP beta=0.8 | max Bellman error | No clear effect |
Closest papers by meaning
- Zap Q-Learning With Nonlinear Function Approximation (NeurIPS 2020 · similarity 0.89)
- Faster Non-asymptotic Convergence for Double Q-learning (NeurIPS 2021 · similarity 0.85)
- A Unified Switching System Perspective and Convergence Analysis of Q-Learning Algorithms (NeurIPS 2020 · similarity 0.85)
- Stabilizing Q-learning with Linear Architectures for Provable Efficient Learning (ICML 2022 · similarity 0.84)
- Linear $Q$-Learning Does Not Diverge in $L^2$: Convergence Rates to a Bounded Set (ICML 2025 · similarity 0.83)
- Quantum Algorithms for Finite-horizon Markov Decision Processes (ICML 2025 · similarity 0.82)
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.
