Data imputation by pursuing better classification: A supervised kernel-based method

Ruikai Yang

doi:10.1016/j.patcog.2025.112312

Data imputation by pursuing better classification: A supervised kernel-based method

2025-08-19

0

PRE

AI

DOI：10.1016/j.patcog.2025.112312

Original

Original request for help

Save

Abstract

En 中文

• We propose a novel two-stage data imputation framework. • We optimize the similarity relationships between data by utilizing supervision information and guide the data imputation process. • We develop a nonparametric method to impute the kernel matrix, which is performed alternately with the training of the classifier. • We provide a solving algorithm based on the block coordinate descent to accurately recover missing data from a given kernel matrix. • When the missing ratio exceeds 60 %, our algorithm outperforms other imputation methods significantly.

Keywords:

data imputation

kernel matrix

nonparametric method

block coordinate descent

similarity optimization

Data imputation by pursuing better classification: A supervised kernel-based method

Abstract

Journal

Organization