Addressing the policy-bias of q-learning by repeating updates.
Sherief AbdallahMichael KaisersPublished in: AAMAS (2013)
Keyphrases
- optimal policy
- action selection
- reinforcement learning
- state space
- policy iteration
- learning algorithm
- cooperative
- function approximation
- markov decision processes
- state action
- reinforcement learning problems
- reinforcement learning algorithms
- reward function
- discounted reward
- multi agent
- evaluation function
- long run
- infinite horizon
- model free
- state dependent
- markov decision process
- least squares
- neural network
- decision problems
- continuous state spaces
- asymptotically optimal
- dynamic environments
- temporal difference learning
- real valued
- decision making
- policy search
- monte carlo
- actor critic
- multi agent reinforcement learning
- stochastic approximation
- rl algorithms
- average cost
- variance reduction
- optimal control
- dynamic programming
- policy makers
- temporal difference
- sufficient conditions