Minimax Number of Strata for Online Stratified Sampling given Noisy Samples - Archive ouverte HAL Access content directly
Preprints, Working Papers, ... Year :

Minimax Number of Strata for Online Stratified Sampling given Noisy Samples

(1) , (1)
1
Alexandra Carpentier
  • Function : Author
  • PersonId : 910455
Rémi Munos
  • Function : Author
  • PersonId : 836863

Abstract

We consider the problem of online stratified sampling for Monte Carlo integration of a function given a finite budget of $n$ noisy evaluations to the function. More precisely we focus on the problem of choosing the number of strata $K$ as a function of the budget $n$. We provide asymptotic and finite-time results on how an oracle that has access to the function would choose the partition optimally. In addition we prove a \textit{lower bound} on the learning rate for the problem of stratified Monte-Carlo. As a result, we are able to state, by improving the bound on its performance, that algorithm MC-UCB, defined in~\citep{MC-UCB}, is minimax optimal both in terms of the number of samples n and the number of strata K, up to a $\sqrt{\log(nK)}$. This enables to deduce a minimax optimal bound on the difference between the performance of the estimate outputted by MC-UCB, and the performance of the estimate outputted by the best oracle static strategy, on the class of Hölder continuous functions, and upt to a $\sqrt{\log(n)}$.
Fichier principal
Vignette du fichier
FixedStrata.pdf (474.05 Ko) Télécharger le fichier
Origin : Files produced by the author(s)
Loading...

Dates and versions

hal-00698517 , version 1 (16-05-2012)

Identifiers

Cite

Alexandra Carpentier, Rémi Munos. Minimax Number of Strata for Online Stratified Sampling given Noisy Samples. 2012. ⟨hal-00698517⟩
177 View
220 Download

Altmetric

Share

Gmail Facebook Twitter LinkedIn More