Importance Sampling: Fundamentals and Applications
Speaker
Time
Friday, November 21, 2025 3:00 PM - 4:00 PM
Venue
A3-2-303
Online
Zoom 435 529 7909
(BIMSA)
Abstract
As a venerable technique in computational statistics, importance sampling finds extensive application in reducing variance and estimating rare-event probabilities. This report will first introduce its theoretical foundations, highlighting the key concept of reweighting important samples to estimate the target expectation. Subsequently, we will illustrate its significant advantages over simple Monte Carlo sampling, particularly its ability to achieve orders of magnitude higher efficiency in challenging scenarios involving tail probabilities and high-dimensional integrals. Finally, the talk will discuss the challenges and practical considerations encountered when applying it to real-world problems.