Overcoming erroneous domain knowledge in plan-based reward shaping.

Kyriakos Efthymiadis Sam Devlin Daniel Kudenko

Published in: AAMAS (2013)

Keyphrases

domain knowledge
reward shaping
reinforcement learning
complex domains
domain theory
background knowledge
reinforcement learning algorithms
domain experts
learning algorithm
steady state
function approximation
markov decision problems
training data
prior knowledge
planning domains