Comparison between the Heston model and the Wasserstein Generative Adversarial Network (WGAN) with gradient penalty.

This work involves the use of a WGAN-GP to model financial stocks, followed by an explanation of its parametrization as well as the Heston Model. Our findings indicate that the WGAN outperforms the Heston Model in terms of kurtosis and skewness :

We also conduct a comparison of trajectory diffusion between the WGAN-GP and the Heston model.

As a group, we decided not to release all of the code we produced for this project. However, we have provided some insights in different notebooks. Please feel free to contact us if you have any comments or questions!