Subwords as Skills: Tokenization for Sparse-Reward Reinforcement Learning.
David YunisJustin JungFalcon Z. DaiMatthew R. WalterPublished in: CoRR (2023)
Keyphrases
- reinforcement learning
- eligibility traces
- function approximation
- state space
- reinforcement learning algorithms
- sparse data
- model free
- reward function
- learning algorithm
- machine learning
- partially observable environments
- supervised learning
- high dimensional
- named entities
- markov decision processes
- biomedical text
- multi agent
- compressive sensing
- temporal difference
- sparse representation
- information extraction
- learning process
- total reward
- biomedical information retrieval
- policy gradient
- lifelong learning
- optimal control
- transfer learning
- dynamic programming
- information retrieval