Deadly triad matters for offline reinforcement learning.
Zhiyong PengYadong LiuZongtan ZhouPublished in: Knowl. Based Syst. (2024)
Keyphrases
- reinforcement learning
- function approximation
- state space
- reinforcement learning algorithms
- robotic control
- learning algorithm
- real time
- model free
- markov decision processes
- temporal difference
- learning problems
- transfer learning
- partially observable
- database
- optimal policy
- evolutionary algorithm
- expert systems
- bayesian networks
- artificial intelligence
- temporal difference learning
- multi agent reinforcement learning
- transition model
- policy search
- neural network
- direct policy search