# Automatic Differentiation

*Quant & Pricing — Finicade finance glossary*

Automatic differentiation computes exact derivatives of a program by applying the chain rule to its operations, rather than by bumping inputs and repricing. In adjoint mode it returns every sensitivity of one output for roughly the cost of a few repricings, regardless of how many inputs there are. For a book with thousands of risk factors that converts an overnight batch into a real-time calculation — the same technique that trains neural networks.

**Also known as:** adjoint algorithmic differentiation, AAD, adjoint greeks

**Related terms:** [The Greeks](https://finicade.com/glossary/the-greeks), [Monte Carlo Simulation](https://finicade.com/glossary/monte-carlo), [Partial Derivative](https://finicade.com/glossary/partial-derivative), [Chain Rule](https://finicade.com/glossary/chain-rule), [Second-Order Greeks](https://finicade.com/glossary/second-order-greeks)

Source: https://finicade.com/glossary/automatic-differentiation