Constrained episodic reinforcement learning in concave-convex and knapsack settings.
Kianté BrantleyMiroslav DudíkThodoris LykourisSobhan MiryoosefiMax SimchowitzAleksandrs SlivkinsWen SunPublished in: NeurIPS (2020)
Keyphrases
- reinforcement learning
- piecewise linear
- dynamic programming
- convexity properties
- convex functions
- convex concave
- knapsack problem
- function approximation
- saddle point
- state space
- convex relaxation
- optimal policy
- learning process
- objective function
- markov decision processes
- convex hull
- data sets
- convex optimization
- multi agent
- reinforcement learning algorithms
- policy search
- feasible solution
- model free
- supervised learning
- np hard
- machine learning