DCT: Dual Channel Training of Action Embeddings for Reinforcement Learning with Large Discrete Action Spaces.
Pranavi PathakotaHardik MeisheriHarshad KhadilkarPublished in: AAMAS (2024)
Keyphrases
- action space
- continuous action
- reinforcement learning
- continuous state
- dual channel
- policy search
- state space
- continuous state spaces
- markov decision processes
- state and action spaces
- real valued
- state action
- action selection
- control policies
- stochastic processes
- reinforcement learning methods
- function approximators
- markov decision process
- function approximation
- supervised learning
- skill learning
- reinforcement learning algorithms
- optimal policy
- partially observable markov decision processes
- infinite horizon
- finite state
- machine learning
- policy iteration
- dynamic programming
- reinforcement learning problems
- model free
- heuristic search