Oracle-SAGE: Planning Ahead in Graph-Based Deep Reinforcement Learning.
Andrew ChesterMichael DannFabio ZambettaJohn ThangarajahPublished in: ECML/PKDD (4) (2022)
Keyphrases
- reinforcement learning
- action selection
- deterministic domains
- planning problems
- function approximation
- database
- partially observable
- reinforcement learning algorithms
- macro actions
- recently developed
- oracle database
- state space
- markov decision processes
- decision support
- graph theoretic
- machine learning
- database administrators
- markov decision problems
- deep learning
- stochastic domains
- robotic control
- blocks world
- single agent
- ai planning
- model free
- neural network
- databases