DCT: Dual Channel Training of Action Embeddings for Reinforcement Learning with Large Discrete Action Spaces.
Pranavi PathakotaHardik MeisheriHarshad KhadilkarPublished in: CoRR (2023)
Keyphrases
- action space
- continuous action
- reinforcement learning
- continuous state
- dual channel
- state space
- markov decision processes
- continuous state spaces
- policy search
- state and action spaces
- real valued
- control policies
- state action
- reinforcement learning methods
- stochastic processes
- action selection
- supervised learning
- reinforcement learning algorithms
- single agent
- learning algorithm
- machine learning
- markov decision process
- policy iteration
- optimal policy
- multi agent
- reward function
- markov decision problems
- dynamic programming
- skill learning
- model free
- probability distribution
- mobile robot
- search space
- multi agent systems