Research map / Reinforcement learning / Policy gradient and actor-critic
Policy Gradient With Value Function Approximation For Collective Multiagent Planning
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 neighborhood count observations (oN) instead of local-state-only observations (o0) (ablation) | Factored actor-critic · Taxi supply-demand matching, 8000 taxis | solution quality: 64% | Helped |
| Use neighborhood count observations (oN) instead of current-state count observations (o1) (ablation) | Factored actor-critic · Taxi supply-demand matching, 8000 taxis | solution quality: 20% | Helped |
| Use current-state count observations (o1) instead of local-state-only observations (o0) (ablation) | Factored actor-critic · Taxi supply-demand matching | solution quality | Helped |
| Use the factored actor and factored critic update (ablation) | Actor-critic · Taxi supply-demand matching, o0, o1, and oN | convergence | Helped |
| Use non-factored actor updates instead of the factored actor update (ablation) | Actor-critic · Taxi supply-demand matching, o0, o1, and oN | convergence | Hurt |
| Use non-factored critic updates instead of the factored critic update (ablation) | Actor-critic · Taxi supply-demand matching, o0, o1, and oN | convergence | Hurt |
| Train the critic with individual value-function targets rather than global empirical returns (ablation) | Actor-critic · Larger collective planning problems | convergence | Helped |
| Use count-dependent policies rather than policies based only on local state (ablation) | Factored actor-critic · Taxi supply-demand matching | solution quality | Helped |
| Use EM with a piecewise policy over neighborhood observations (ablation) | EM · Taxi supply-demand matching, oN | convergence: 30K iterations | Hurt |
| Use the factored actor-critic method instead of SMFU | Collective multiagent planning · Robot navigation in a congested environment, oN | solution quality | Helped |
| Use neighborhood count observations (oN) instead of less informative observation models (ablation) | Factored actor-critic · Robot navigation in a congested environment | solution quality | Helped |
Closest papers by meaning
- Credit Assignment For Collective Multiagent RL With Global Rewards (NeurIPS 2018 · similarity 0.89)
- Learning to Act in Decentralized Partially Observable MDPs (ICML 2018 · similarity 0.89)
- Collective Multiagent Sequential Decision Making Under Uncertainty (AAAI 2017 · similarity 0.89)
- Multi-Agent Reinforcement Learning with General Utilities via Decentralized Shadow Reward Actor-Critic (AAAI 2022 · similarity 0.89)
- Learning Decentralized Partially Observable Mean Field Control for Artificial Collective Behavior (ICLR 2024 · similarity 0.86)
- One Step is Enough: Multi-Agent Reinforcement Learning Based on One-Step Policy Optimization for Order Dispatch on Ride-Sharing Platforms (NeurIPS 2026 · similarity 0.86)
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.