This chapter studies algorithms with polynomial sample complexity of exploration, both model-based and model-free ones, in a unified manner. These so-called PAC ...

Abstract. Efficient exploration is widely recognized as a fundamental challenge in- herent in reinforcement learning. Algorithms that explore efficiently ...

We design new algorithms for the combinatorial pure exploration problem in the multi-arm bandit framework.

Nov 21, 2017 · We design new algorithms for the combinatorial pure exploration problem in the multi-arm bandit framework.

The bounds derived there can also be used to derive a lower bound for our problem, but do not appear to be tight enough to capture the log(1/δ) dependence on δ.

sample complexity bounds are better for model-free methods than for the model-based ones: the sample complexity of delayed Q-learning scales linearly with ...

Sample Complexity Bounds and an Efficient Algorithm. Tongyi Cao and ... Nearly optimal sampling algorithms for combinatorial pure exploration. COLT ...

We establish a matching lower bound on the expected number of trials under any sampling policy. We furthermore generalize the lower bound, and show an ...

Missing: Exploration. | Show results with:Exploration.

It aims to reduce the sample complexity of reinforcement learning by collecting the right data. ... Li: Sample complexity bounds of exploration. In Marco Wiering ...