Interpretable policies for reinforcement learning by genetic programming.
Daniel HeinSteffen UdluftThomas A. RunklerPublished in: Eng. Appl. Artif. Intell. (2018)
Keyphrases
- genetic programming
- reinforcement learning
- optimal policy
- classification rules
- policy search
- control policies
- fitness function
- evolutionary algorithm
- state space
- markov decision process
- fitted q iteration
- evolutionary computation
- symbolic regression
- policy gradient methods
- control policy
- reinforcement learning algorithms
- total reward
- markov decision problems
- markov decision processes
- model free
- gene expression programming
- genetic algorithm
- reward function
- function approximation
- decision problems
- partially observable markov decision processes
- dynamic programming
- reinforcement learning agents
- macro actions
- average cost
- long run
- financial forecasting
- grammar guided genetic programming
- function approximators
- finite state
- average reward
- learning algorithm
- temporal difference
- hierarchical reinforcement learning
- approximate policy iteration
- learning process
- machine learning
- continuous state
- transfer learning
- temporal difference learning
- optimal control
- regression problems
- state abstraction
- multi agent reinforcement learning
- management policies
- policy gradient
- evolutionary approaches
- sufficient conditions
- reinforcement learning methods
- multi agent
- robotic control
- action selection