Integrated Imputation-Classification for Supervised Learning with Missing Data

2026-10-03

AI Summary

This study addresses the loss of predictive information caused by decoupling imputation and classification in supervised learning with missing values. To this end, we propose the IICN network, which introduces an additional "imputed" class to construct an (n+1)-class discriminator. Through adversarial training, the imputer and discriminator are jointly optimized, enabling end-to-end classification on incomplete data. Theoretically, we prove that at global optimality, the model performs marginalization and achieves Bayes optimality. Experimental results demonstrate that IICN significantly outperforms conventional pipeline approaches and generative baselines on datasets such as FashionMNIST, effectively improving both classification accuracy and robustness.

Abstract

We study supervised classification problems with missing feature values. Existing approaches often decouple imputation from classification, producing imputations that may be plausible but uninformative for prediction. Instead, we propose the Integrated Imputation and Classification Network (IICN), which jointly trains an imputer and an ${(n{+}1)}$-classdiscriminator adversarially with a single class supervised classification objective, where the discriminator learns to distinguish among the $n$ true classes and an additional ``imputed"class. We prove that at the global optimum, the imputer and discriminator together implement marginalization over missing coordinates and yield a Bayes-optimal classifier. We evaluate IICN on FashionMNIST, CIFAR-10, and tabular datasets with naturally occurring missingness. IICN outperforms classical impute-then-classify pipelines and recent generative baselines, showing strong robustness and accuracy in challenging settings.

Problem

• supervised classification
• missing data
• imputation
• feature values

Innovation

• Integrated Imputation-Classification
• Adversarial Training
• Missing Data
• (n+1)-class Discriminator
• Bayes-optimal Classifier