No-Regret Reinforcement Learning in Smooth MDPs.
Davide MaranAlberto Maria MetelliMatteo PapiniMarcello RestelliPublished in: CoRR (2024)
Keyphrases
- reinforcement learning
- markov decision processes
- reward function
- total reward
- reinforcement learning algorithms
- state space
- optimal policy
- online learning
- function approximation
- markov decision process
- learning algorithm
- policy search
- control problems
- multi armed bandit
- reinforcement learning methods
- partially observable
- lower bound
- state and action spaces
- dynamic programming
- temporal difference
- multi agent
- machine learning
- policy iteration
- continuous state and action spaces
- action sets
- action space
- expert advice
- confidence bounds
- expected reward
- transition model
- continuous state
- markov decision problems
- average reward
- markov chain
- finite state
- model free
- infinite horizon
- approximate dynamic programming
- continuous state spaces
- worst case
- finite horizon
- loss function
- state variables
- model based reinforcement learning