Enhanced Importance Sampling Through Latent Space Exploration in Normalizing Flows
Liam Kruse, Alexandros Tzikas, Harrison Delecki, Mansur Arief, Mykel Kochenderfer · 2025
Read Enhanced Importance Sampling Through Latent Space Exploration in Normalizing Flows by Liam Kruse, Alexandros Tzikas, Harrison Delecki, Mansur Arief, Mykel Kochenderfer (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2501-03394.txt.
Description
Liam Kruse and four co-authors hunt rare events in a model's dream space in a 2025 paper. Importance sampling speeds up estimating rare disasters by drawing extra samples near the danger and reweighting. It fails when the drawing plan misses where the danger truly lives.
Their plan searches inside a normalizing flow, a learned map that unfolds tangled data into a tidy latent ball. The latent space is easy to explore, and the map folds any proposal back into real samples exactly. Tests on simulated racing cars and aircraft collision avoidance showed sharper rare-event estimates. The trick matters wherever the rare case is the one that kills.
Catalog record
| CREATOR | Liam Kruse, Alexandros Tzikas, Harrison Delecki, Mansur Arief, Mykel Kochenderfer |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 31 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2501-03394