Mitigating Estimation Errors by Twin TD-Regularized Actor and Critic for Deep Reinforcement Learning.
Junmin ZhongRuofan WuJennie SiPublished in: CoRR (2023)
Keyphrases
- estimation error
- temporal difference
- reinforcement learning
- reinforcement learning algorithms
- step size
- td learning
- function approximation
- actor critic
- temporal difference learning
- model free
- evaluation function
- eligibility traces
- estimation accuracy
- policy evaluation
- state space
- function approximators
- convergence rate
- temporal difference methods
- monte carlo
- error rate
- least squares
- covariance matrix
- markov decision processes
- standard deviation
- learning algorithm
- action selection
- policy iteration
- convergence speed
- cost function
- reinforcement learning problems
- fourth order
- supervised learning
- multi agent
- reinforcement learning methods
- gradient method
- reward function
- objective function
- learning tasks
- machine learning