Conditional GAN Approaches on Regression Labels: A State‐of‐the‐Art Review

Abstract

This paper presents a comprehensive review of the most popular algorithms available nowadays, for the generation of synthetic data guided by continuous labeling, based on Generative Adversarial Networks (GANs). It is well known that GANs have produced an outbreak in Artificial Intelligence, particularly in deep learning (DL), where the research on models capable of generating realistic data grows daily. However, the work currently developed related to data generation driven by regression labels is rather modest, even though the number of applications is enormous, which makes it mandatory to intensify the research related to this area. Here, we classify and discuss several continuous GAN models (cGANs), methodologies, and applications currently available, showing some of their success areas, as well as the principal challenges found during their practical use. This article is categorized under: Technologies > Machine Learning Application Areas > Science and Technology Technologies > Computational Intelligence