PoliFormer: Scaling On-Policy RL with Transformers Results in Masterful Navigators.
Kuo-Hao ZengZichen ZhangKiana EhsaniRose HendrixJordi SalvadorAlvaro HerrastiRoss GirshickAniruddha KembhaviLuca WeihsPublished in: CoRR (2024)
Keyphrases
- optimal policy
- reinforcement learning
- action selection
- control policy
- markov decision process
- control policies
- policy gradient
- markov decision processes
- action space
- policy iteration
- model free
- policy search
- partially observable domains
- state space
- approximate dynamic programming
- dynamic programming
- actor critic
- function approximation
- long run
- policy evaluation
- state and action spaces
- state action
- partially observable markov decision processes
- markov decision problems
- reinforcement learning problems
- agent learns
- exploration exploitation tradeoff
- average reward
- genetic algorithm
- partially observable
- reinforcement learning algorithms
- multi agent
- decision making
- learning algorithm
- asymptotically optimal
- sequential decision making
- temporal difference
- infinite horizon
- partially observable markov decision process
- finite state
- conflict resolution
- learning process
- discounted reward