Bridging the Gap Between Value and Policy Based Reinforcement Learning.
Ofir NachumMohammad NorouziKelvin XuDale SchuurmansPublished in: NIPS (2017)
Keyphrases
- reinforcement learning
- optimal policy
- policy search
- markov decision process
- action selection
- actor critic
- partially observable environments
- reinforcement learning problems
- policy gradient
- reinforcement learning algorithms
- policy evaluation
- partially observable
- function approximation
- markov decision problems
- partially observable markov decision processes
- control policy
- markov decision processes
- action space
- reward function
- function approximators
- approximate dynamic programming
- state space
- control policies
- state and action spaces
- temporal difference
- continuous state spaces
- policy iteration
- state action
- rl algorithms
- decision problems
- partially observable domains
- policy gradient methods
- average reward
- temporal difference learning
- learning algorithm
- dynamic programming
- infinite horizon
- neural network
- agent learns
- multi agent reinforcement learning
- long run
- model free
- state dependent
- learning classifier systems
- human users
- optimal control
- transfer learning
- supervised learning
- policy making
- asymptotically optimal
- agent receives