Learning General Optimal Policies with Graph Neural Networks: Expressive Power, Transparency, and Limits.
Simon StåhlbergBlai BonetHector GeffnerPublished in: CoRR (2021)
Keyphrases
- expressive power
- optimal policy
- reinforcement learning
- first order logic
- learning algorithm
- transitive closure
- query language
- special case
- relational algebra
- average reward reinforcement learning
- state space
- dynamic programming
- markov decision processes
- long run
- finite horizon
- relational calculus
- serial inventory systems
- dynamic programming algorithms
- data complexity
- database systems