查看文章

Generative adversarial networks (GANs) are capable of learning deep representations efficiently without requiring a significant amount of annotation. Through Backpropagation, signals are derived from one network by another through a competitive process. This article presents a machine learning framework that incorporates generative adversarial networks (GANs) and interactive evolutionary computation (IEC) to generate images and to resample those images for the detection of fake images. In our study, we found that GANs trained on a particular domain could produce reliable and compact phenotypic maps. Through the use of a user research method, participants able to produce images closely matched target images, demonstrating the advantages of this unique approach.