Model Calibration with Neural Networks

Abstract

A central consideration for the use of any pricing model is the ability to calibrate that model to market or historical prices. Whether the information needed by the model can be effectively implied from the data or not is one part of the calibration problem. However, in many applications, the speed with which that calibration can be performed influences the usability of that model.

In the following a method is presented to calibrate models using artificial neural networks, which can perform the calibration significantly faster regardless of the model, hence removing the calibration speed from consideration for a model's usability.

Keywords: model calibration, neural networks

JEL Classification: G10, G13, C45, E47