artificial-intelligence
近端策略优化
Proximal Policy OptimizationA model-free, on-policy reinforcement learning algorithm that optimizes a policy by limiting the size of policy updates to ensure stability and prevent divergence. / 一种无模型、在线策略的强化学习算法，通过限制策略更新的幅度来确保稳定性和防止发散。
- Knowledge type
- 数学理论
- First proposed or discovered
- 2017
- Proposer
- John Schulman
- Creator
- 农场主
- Last updated
- 2026-11-03 19:06:47
Sub-knowledge points
Actor-Critic architectureTrust region policy optimizationAdvantage actor-critic (A2C)
Higher-order knowledge points
Main parameters and performance indicators
learning rate: step size for policy and value function updatesclip range: threshold for policy update ratio to prevent large changesdiscount factor: weight for future rewards
Open problems
sample efficiencyexploration-exploitation trade-offscalability to high-dimensional continuous action spaces
Related literature
ranked by community starsMulti-head actor-critic proximal policy optimization for coordinated control of EV charging and battery storage in PV-integrated microgridsOpen source
Related products and links
No item recorded yet
Creation and modification trail
1 recordsRecord title: Proximal Policy Optimization.
Node update feedback
review required