Knoop : Practical Enhancement of Knockoff with Over-Parameterization for Variable Selection

Xiaochen Zhang, Yunfeng Cai, Haoyi Xiong · 2025

Read Knoop : Practical Enhancement of Knockoff with Over-Parameterization for Variable Selection by Xiaochen Zhang, Yunfeng Cai, Haoyi Xiong (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-2501-17889.txt.

Description

Xiaochen Zhang, Yunfeng Cai, and Haoyi Xiong put forth Knoop in 2025, a way to pick which inputs truly matter in a model. The base method, called knockoff, makes fake twin inputs and keeps real ones that beat their twins. Knoop makes many twins per input and feeds them all into a spare model with no slack. It then checks how each set of twin weights spreads.

A real input whose twins spread in an odd way stands out as a true signal. Tests on made-up and real data showed strong gains over past methods. Scores for finding the true inputs rose, and forecast accuracy rose across tasks that sort and that grade. The authors back the gains with math results as well.

Catalog record

CREATORXiaochen Zhang, Yunfeng Cai, Haoyi Xiong
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE58 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2501-17889

Browse the full library · About the Tower of Babel