Artificial Intelligence
深度确定性策略梯度
Deep Deterministic Policy GradientA model-free, off-policy actor-critic algorithm for continuous action control problems / 一种用于连续动作控制问题的无模型、离策略演员-评论家算法
- Knowledge type
- 数学理论
- First proposed or discovered
- 2016
- Proposer
- Lillicrap et al.
- Creator
- 农场主
- Last updated
- 2026-10-28 22:00:25
Sub-knowledge points
Actor-Critic architectureDeterministic policySoft update mechanism
Higher-order knowledge points
Main parameters and performance indicators
actor network: approximates the deterministic policy functioncritic network: approximates the Q-value functionreplay buffer: stores transitions for off-policy learningtarget networks: stabilize training by providing stable targets
Open problems
Exploration in continuous action spacesSample efficiencyStability during training
Related literature
ranked by community starsEnergy-efficient routing scheme integrating small-world network with n-step DDQN for underwater wireless sensor networksOpen source
Related products and links
No item recorded yet
Creation and modification trail
1 recordsRecord title: Deep Deterministic Policy Gradient.
Node update feedback
review required