This thesis provides a comprehensive survey of the mathematical theory underlying Generative Adversarial Networks (GANs), exploring the algorithm through three complementary theoretical frameworks: game theory, information theory, and optimal transport theory.