Toward Fusing Domain Knowledge with Generative Adversarial Networks to Improve Supervised Learning for Medical Diagnoses

Fu-Chieh Chang, Jocelyn Chang, Chun-Nan Chou, Edward Y. Chang

Abstract: This paper addresses the challenges of small training data in deep learning. We share our experiences in the medical domain and present promises and limitations. In particular, we show through experimental results that GANs are ineffective in generating quality training data to improve supervised learning. We suggest plausible research directions to remedy the problems.