Interpretable Policies for Reinforcement Learning by Genetic Programming.
Daniel HeinSteffen UdluftThomas A. RunklerPublished in: CoRR (2017)
Keyphrases
- robotic control
- genetic programming
- reinforcement learning
- classification rules
- optimal policy
- fitness function
- evolutionary algorithm
- evolutionary computation
- markov decision process
- gene expression programming
- symbolic regression
- policy search
- regression problems
- financial forecasting
- hierarchical reinforcement learning
- reinforcement learning agents
- grammar guided genetic programming
- control policies
- learning algorithm
- reward function
- decision problems
- partially observable markov decision processes
- markov decision processes
- control policy
- markov decision problems
- approximate policy iteration
- genetic algorithm
- fitted q iteration
- temporal difference
- state space
- partially observable
- reinforcement learning algorithms
- evolutionary approaches
- model free
- finite state
- total reward
- optimal control
- function approximation
- policy gradient methods
- dynamic programming