An Exponential Lower Bound for Linearly Realizable MDP with Constant Suboptimality Gap.
Yuanhao WangRuosong WangSham M. KakadePublished in: NeurIPS (2021)
Keyphrases
- lower bound
- upper bound
- vc dimension
- markov decision processes
- average case complexity
- branch and bound algorithm
- constant factor
- lower and upper bounds
- optimal solution
- branch and bound
- np hard
- arbitrarily close
- utility function
- optimal policy
- linear program
- dynamic programming
- lower bounding
- linear programming relaxation
- objective function
- markov decision process
- sufficient conditions
- reinforcement learning
- multi agent
- online algorithms
- heuristic search
- sufficiently accurate
- state space