Reinforcement Learning as Heuristic for Action-Rule Preferences.
Joost BroekensKoen V. HindriksPascal WiggersPublished in: ProMAS (2010)
Keyphrases
- reinforcement learning
- action selection
- action descriptions
- dynamic programming
- action space
- partially observable domains
- reward shaping
- action rules
- condition action rules
- optimal solution
- state action
- reinforcement learning algorithms
- state space
- user preferences
- function approximation
- decision making
- transition model
- learning algorithm
- markov decision processes
- genetic algorithm
- multi agent
- sensory inputs
- association rules
- tie breaking
- agent learns
- machine learning
- search algorithm
- preference elicitation
- hyper heuristics
- temporal difference
- rule sets
- tabu search
- optimal policy
- reasoning about actions
- soft constraints
- domain description
- human actions
- learning process