Explaining Deep Reinforcement Learning Policies with SHAP, Decision Trees, and Prototypes.
Kristoffer EiksåJan Erlend VatneAnastasios M. LekkasPublished in: MED (2024)
Keyphrases
- decision trees
- reinforcement learning
- optimal policy
- policy search
- markov decision process
- control policies
- training set
- function approximation
- hierarchical reinforcement learning
- state space
- machine learning
- fitted q iteration
- reinforcement learning agents
- markov decision processes
- reinforcement learning algorithms
- reward function
- markov decision problems
- partially observable markov decision processes
- naive bayes
- policy gradient methods
- control policy
- decision tree induction
- multi agent
- decision problems
- rule sets
- predictive accuracy
- neural network
- model free
- finite state
- random forest
- dynamic programming
- long run
- rule induction
- machine learning algorithms
- learning process
- deep learning
- learning algorithm
- decision rules
- partially observable
- feature construction
- classification rules
- sufficient conditions
- supervised learning
- feature selection
- data mining methods
- optimal control
- learning classifier systems
- infinite horizon