Natural Language Specification of Reinforcement Learning Policies Through Differentiable Decision Trees.
Pradyumna TambwekarAndrew SilvaNakul GopalanMatthew C. GombolayPublished in: IEEE Robotics Autom. Lett. (2023)
Keyphrases
- decision trees
- reinforcement learning
- natural language
- optimal policy
- machine learning
- policy search
- control policies
- markov decision process
- formal language
- markov decision processes
- reward function
- function approximation
- objective function
- fitted q iteration
- partially observable markov decision processes
- reinforcement learning agents
- formal languages
- state space
- reinforcement learning algorithms
- information extraction
- decision tree induction
- semantic interpretation
- random forest
- natural language interface
- policy gradient methods
- natural language processing
- semantic analysis
- knowledge representation
- markov decision problems
- naive bayes
- dialogue system
- decentralized control
- dynamic programming
- natural language generation
- machine learning algorithms
- decision problems
- action selection
- continuous state
- finite state
- data mining methods
- predictive accuracy
- hierarchical reinforcement learning
- constructive induction
- learning algorithm
- control policy
- temporal difference
- formal specification
- question answering
- training data
- action space
- feature construction
- decision tree learning
- language processing
- infinite horizon
- multi agent reinforcement learning
- optimal control
- loss function
- average reward
- multi agent
- high level