Generative Adversarial NetworksAdversarialgenerative adversarial network (GAN) (GANs)Generativegenerative adversarial network (GAN) are a classClass of generative modelsGenerativegenerative model where two networks are trained together: a generator that produces synthetic samples from noise, and a discriminator that attempts to distinguish real samples from synthetic ones. Through this adversarial gameAdversarialadversarial game, the generator gradually improves at producing realistic outputs. This adversarial process enables the generator to enhance the quality of generated samples without relying on human judgment, making GANsAdversarialgenerative adversarial network (GAN) highly effective for tasksTask such as image synthesis, data augmentation, and beyond. This chapter provides a comprehensive introduction toGenerativegenerative adversarial network (GAN) GANsAdversarialgenerative adversarial network (GAN) and their underlying principles. Key challenges, such asModemode collapse mode collapseCollapsemode collapse, are discussed alongside theoretical perspectives including maximum likelihoodLikelihood estimation withinGenerativegenerative adversarial network (GAN) GANsAdversarialgenerative adversarial network (GAN). The chapter also surveys important GAN variants, sampling and interpolation techniques, and highlights diverse applications. Finally, it explores extensionsExtension of adversarial learningAdversarialadversarial learning within autoencoder architecturesArchitecture. The content aims to equip readers with a foundational understanding as well as insights into recent developments and practicalPracticepractical applications of adversarial generative modeling.