FragmGAN: Generative Adversarial Nets for Fragmentary Data Imputation and Prediction
This addresses the challenge of fragmentary data in scientific research and applications, offering a unified solution with theoretical guarantees for MAR data, though it appears incremental as it builds on existing GAN methods.
The authors tackled the problem of imputation and prediction with fragmentary data by proposing FragmGAN, a GAN-based framework that simultaneously handles imputation and label prediction, achieving significant advantages in predictive performance in experiments.
Modern scientific research and applications very often encounter "fragmentary data" which brings big challenges to imputation and prediction. By leveraging the structure of response patterns, we propose a unified and flexible framework based on Generative Adversarial Nets (GAN) to deal with fragmentary data imputation and label prediction at the same time. Unlike most of the other generative model based imputation methods that either have no theoretical guarantee or only consider Missing Completed At Random (MCAR), the proposed FragmGAN has theoretical guarantees for imputation with data Missing At Random (MAR) while no hint mechanism is needed. FragmGAN trains a predictor with the generator and discriminator simultaneously. This linkage mechanism shows significant advantages for predictive performances in extensive experiments.