Provable Benefits of Representational Transfer in Reinforcement Learning.
Alekh AgarwalYuda SongWen SunKaiwen WangMengdi WangXuezhou ZhangPublished in: CoRR (2022)
Keyphrases
- reinforcement learning
- transfer learning
- function approximation
- machine learning
- real time
- learning process
- control problems
- robot control
- model free
- markov decision processes
- state space
- learning algorithm
- information retrieval
- real world
- neural network
- action space
- multiple representations
- markov decision process
- reinforcement learning methods
- previously learned
- reinforcement learning algorithms
- optimal policy
- dynamic programming
- optimal solution
- multi agent
- search engine