Differential machine learning technique expanded for derivative pricing

Key facts

• Differential machine learning trains approximations for complex pricing models by targeting price sensitivities.
• The original method was limited by requiring payout continuity.
• The limitation excluded barrier and digital features in derivative pricing.
• Paul Glasserman and Siddharth Hemant Karmarkar developed an expanded scope for the method.

A technique known as differential machine learning, which trains fast approximations for complex pricing models by targeting price sensitivities, has been expanded. Previously, this method required payout continuity, which ruled out the inclusion of barrier and digital features in derivative pricing. Paul Glasserman and Siddharth Hemant Karmarkar have now broadened the scope of this method to eliminate this limitation, thereby increasing its applicability to a wider range of derivative pricing scenarios.