Policy Gradient Methods
Twin Delayed Deep Deterministic
Introduced by Fujimoto et al. in Addressing Function Approximation Error in Actor-Critic MethodsTD3 builds on the DDPG algorithm for reinforcement learning, with a couple of modifications aimed at tackling overestimation bias with the value function. In particular, it utilises clipped double Q-learning, delayed update of target and policy networks, and target policy smoothing (which is similar to a SARSA based update; a safer update, as they provide higher value to actions resistant to perturbations).
Source: Addressing Function Approximation Error in Actor-Critic Methods
Papers
| Paper | Code | Results | Date | Stars |
|---|
Tasks
| Task | Papers | Share |
|---|---|---|
| Reinforcement Learning (RL) | 58 | 40.56% |
| Continuous Control | 26 | 18.18% |
| OpenAI Gym | 8 | 5.59% |
| Decision Making | 7 | 4.90% |
| Autonomous Driving | 5 | 3.50% |
| Offline RL | 3 | 2.10% |
| Meta-Learning | 3 | 2.10% |
| Benchmarking | 3 | 2.10% |
| D4RL | 2 | 1.40% |
Adam
Clipped Double Q-learning
Dense Connections
Experience Replay
ReLU
Target Policy Smoothing
