What question did this study set out to answer?

To develop and evaluate a power-based regularization strategy for adaptive importance sampling that mitigates high weight variance and improves target distribution estimates.

October 29, 2021Open Access

Adaptive Importance Sampling meets Mirror Descent: a Bias-variance tradeoff

View Full Paper

Ask AI

Bookmark

Authors

AKAnna KorbaTélécom Paris FPFrançois PortierÉcole Nationale de la Statistique et de l'Analyse de l'Information

Discussion

Member takes

Implication

Algorithmic analysis demonstrates that power-regularized weights manage the bias-variance tradeoff in adaptive importance sampling, indicating improved estimation accuracy via mirror descent.

Key Points

To develop and evaluate a power-based regularization strategy for adaptive importance sampling that mitigates high weight variance and improves target distribution estimates.
Introduced a regularization mechanism that raises importance weights to a power parameter between zero and one.
Framed the regularization parameter update within the mirror descent optimization framework.
Formulated the adaptive sampling policy using kernel density estimation and analyzed uniform convergence conditions.
Demonstrated that the power parameter controls an explicit bias-variance tradeoff in iterative importance sampling.
Established theoretical uniform convergence of the kernel density sampling policy under mild assumptions.
Confirmed through empirical experiments that regularized importance weighting yields superior estimation performance compared to standard adaptive sampling.

Cite This Study

Korba et al. (2021) studied this question.

synapsesocial.com/papers/6a1e983fd258f1f64202b4dd https://doi.org/10.48550/arxiv.2110.15590

View Full Paper

Ask AI

Bookmark

Implication

Algorithmic analysis demonstrates that power-regularized weights manage the bias-variance tradeoff in adaptive importance sampling, indicating improved estimation accuracy via mirror descent.

Adaptive Importance Sampling meets Mirror Descent: a Bias-variance tradeoff

Discussion

Member takes

Implication

Key Points

Cite This Study

Adaptive Importance Sampling meets Mirror Descent: a Bias-variance tradeoff

Discussion

Member takes

Implication

Key Points

Cite This Study

Also Consider

Also Consider