The project aims to model the Stochastic Discount Factor (SDF) in asset pricing using a Generative Adversarial Network (GAN). The approach allows the generator to derive an SDF conditioned on the state of the economy which is dynamically captured through macroeconomic time-series data processed by an LSTM. Concurrently, an adversarial Discriminator network generates optimal test portfolios specifically designed to maximize the model's out-of-sample pricing errors.