Research map / Learning theory and algorithms / Combinatorial and search optimization
Fast Approximate Dynamic Programming for Infinite-Horizon Markov Decision Processes
Mohamad Amin Sharifi Kolarijani, Gyula Felix Max, Peyman Mohajerin Esfahani
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 ConjVI instead of primal VI | value iteration · Synthetic stochastic system, N=41 | iterations to termination: 102 -> 55 | Improved |
| use dynamic dual grid (ConjVI-d) instead of static dual grid (ablation) | ConjVI · Synthetic stochastic system | iterations to termination: 55 -> 100 | Hurt |
| use dynamic dual grid (ConjVI-d) instead of static dual grid (ablation) | ConjVI · Synthetic stochastic system | greedy policy cost | Improved |
| use ConjVI instead of primal VI | value iteration · Synthetic, inverted pendulum | greedy policy average cost | Hurt |
| use ConjVI instead of primal VI | value iteration · Batch reactor eta=0.01 | run-time (sec): 7669 -> 55 | Improved |
| use ConjVI instead of primal VI | value iteration · Batch reactor eta=0.01 | average cost (100 runs): 33.9 -> 73.5 | Hurt |
| use ConjVI-d instead of primal VI when cost has large Lipschitz constant | value iteration · Batch reactor eta=0.01 | average cost (100 runs): 33.9 -> 74.0 | Hurt |
| use dynamic dual grid instead of static dual grid (ablation) | ConjVI · Batch reactor eta=0.01 | run-time (sec): 55 -> 90 | Hurt |
| use dynamic dual grid instead of static dual grid (ablation) | ConjVI · Batch reactor (default cost) | greedy policy cost | Improved |
| deterministic dynamics instead of stochastic | ConjVI-d · Synthetic system | iterations to termination: 100 -> 10 | Improved |
| use analytic conjugate of input cost instead of numerical approximation (ablation) | ConjVI · Synthetic system | value function and iterations | No clear effect |
| scaling factor alpha = 3 for dynamic dual grid (ablation) | ConjVI-d · Inverted pendulum | convergence | Hurt |
Closest papers by meaning
- Rank-One Modified Value Iteration (ICML 2025 · similarity 0.86)
- Augmented Lagrangian Method for Last-Iterate Convergence for Constrained MDPs (NeurIPS 2026 · similarity 0.85)
- Sketched Newton Value Iteration for Large-Scale Markov Decision Processes (AAAI 2024 · similarity 0.85)
- Truncated Variance Reduced Value Iteration (NeurIPS 2024 · similarity 0.85)
- Efficient Value Iteration for s-rectangular Robust Markov Decision Processes (ICML 2024 · similarity 0.84)
- Faster Fixed-Point Methods for Multichain MDPs (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.
