Generative adversarial networks for multi-fidelity matrix completion with massive missing entries

Zongqi Liu, Xueguan Song, Jie Yang, Chao Zhang, Dacheng Tao

Abstract: Highlights•An incomplete HF matrix is completed by only using very few available HF entries and an LF matrix.•A generative adversarial network is designed for solving multi-fidelity matrix completion problem.•A matrix data augmentation method is adopted to enrich the samples for training the network.•The correlation coefficient is used to describe the relatedness between LF and HF matrices.•GAN-MMC performs well even when the available HF entries are rarely few.