ASSET PRICING OPTIMIZATION THROUGH GENERATIVE ADVERSARIAL NETWORKS

Abstract

In this thesis, we study the asset pricing optimisation through Generative Adversarial Networks (GAN). We have demonstrated that shallow learning can deliver similar performance for test data as compared to deep learning considered in the literature, with the added benefit of mitigating common challenges such as overfitting. This is an important finding, as it challenges the often-held belief that more complex, deeper models are invariably superior. Instead, we found that a simpler, less computationally intensive model can provide comparable results, and potentially do so with greater efficiency. While deep learning certainly has its merits, especially for more complex tasks, our work underlines the significance of model appropriateness and the trade-offs between model complexity and performance. The issue of overfitting, which is often more pronounced in deeper networks, has been less problematic in our shallow learning approach.