Parametric Differential Machine Learning for Pricing and Calibration

Graph shows different types of financial data


We developed a Parametric Differential Machine Learning methodology to learn Deep Neural Network parametric pricers for varying model and contract parameters, with adaptive parametric sampling. We demonstrated these parametric pricers and used them for calibration for the example of Cheyette models for interest rate caplets. We used the inherent randomness of the process to optimize over several random replications and thus robustify the calibration. Models and instruments are given in low-code close to mathematical notation and then translated to efficient differentiable simulation and computation in TensorFlow. This is joint work with Bernhard Hientzsch.


Arun Polala is a senior quantitative analyst at Wells Fargo Bank. He graduated from Florida State University with a Ph.D. degree in Financial Mathematics. As a senior quantitative analyst, he is responsible for developing financial models for pricing and risk management using advanced Machine Learning techniques.

Zoom link