Policy Gradient Methods
ACER
Introduced by Wang et al. in Sample Efficient Actor-Critic with Experience ReplayACER, or Actor Critic with Experience Replay, is an actor-critic deep reinforcement learning agent with experience replay. It can be seen as an off-policy extension of A3C, where the off-policy estimator is made feasible by:
- Using Retrace Q-value estimation.
- Using truncated importance sampling with bias correction.
- Using a trust region policy optimization method.
- Using a stochastic dueling network architecture.
Papers
| Paper | Code | Results | Date | Stars |
|---|
Tasks
| Task | Papers | Share |
|---|---|---|
| Reinforcement Learning (RL) | 5 | 26.32% |
| Face Anti-Spoofing | 3 | 15.79% |
| Face Recognition | 3 | 15.79% |
| Problem Decomposition | 2 | 10.53% |
| Face Presentation Attack Detection | 1 | 5.26% |
| Automatic Speech Recognition (ASR) | 1 | 5.26% |
| Benchmarking | 1 | 5.26% |
| Speech Recognition | 1 | 5.26% |
| Spoken Dialogue Systems | 1 | 5.26% |
Convolution
Dense Connections
Entropy Regularization
Experience Replay
ReLU
Retrace
Softmax
Stochastic Dueling Network
TRPO
