Approximate Dynamic Programming with Neural Networks in Linear Discrete Action Spaces.
Wouter van HeeswijkHan La PoutréPublished in: CoRR (2019)
Keyphrases
- approximate dynamic programming
- reinforcement learning
- action space
- function approximators
- continuous action
- continuous state
- markov decision processes
- policy iteration
- linear program
- dynamic programming
- policy search
- state space
- average cost
- step size
- function approximation
- real valued
- control policy
- optimal policy
- linear programming
- stochastic processes
- optimal control
- finite state
- temporal difference
- action selection
- partially observable markov decision processes
- model free
- data mining
- multistage
- probabilistic model
- multi agent
- learning algorithm
- genetic algorithm