Tecniche Avanzate nella Calibrazione della Volatilità Stocastica

Techniques from deep learning play an increasingly impor- tant role in the calibration of financial models. Instead of de- riving model parameters directly from observed market data, this approach utilizes a neural-network-based approximating pricing map. With this efficient map established, conventional calibra- tion algorithms can be subsequently employed. The Black and Sc- holes model, a cornerstone in finance, has its limitations, prompt- ing the development of models like SABR, Heston, and Hull and White. Recent attention has been directed towards rough stochas- tic volatility models. In this context, the primary focus of this research is to leverage neural networks to discern the implied volatility (IV) map, studying a more efficient alternative to the traditionally resource-intensive Monte Carlo simulations used in calibration tasks. By analyzing BTC option data, this study un- derscores the adaptability and efficacy of deep learning techniques in refining financial model calibration.