A Review of Generative Adversarial Networks for Addressing Data Imbalance and Enhancing Model Performance

Abstract

In current years, Generative Adversarial Networks (GANs) have grown into a popular and active research domain in artificial intelligence and machine learning. GANs provide an adequate data augmentation method, which helps improve model accuracy by generating realistic synthetic samples, especially for underrepresented classes. This paper reviews various techniques and methods used in training GANs, particularly focusing on their role in addressing dataset imbalance. The paper also discusses the implications of utilizing GANs for improving model generalization, mitigating bias, and reducing overfitting. This paper discusses the training methods for generative adversarial models, highlighting their significance. It overviews various model design strategies, algorithms, and recent approaches to enhance training.