Learning Distributions for Continuous-Time Financial Models

This study introduces a novel approach that uses neural networks to efficiently compute distributions of continuous-time financial models, bypassing the need for explicit SDE solutions and numerical methods like Monte Carlo simulation. Our approach mitigates challenges that employing the traditional methods incur high computational costs by training a neural network to approximate the empirical cumulative distribution function from the Monte Carlo simulation based on the Glivenko-Cantelli theorem. This approach not only enhances computational efficiency but also provides a viable tool for pricing options, demonstrating significant improvements over traditional methods in both speed and accuracy.