---

4 Functions for numerical approximation of gradients and Hessians ¶

You should not usually need to use these functions directly or pass them as arguments. They should be chosen and used by optimizer functions, possibly subject to their configuration options.

• Direct user interface to default numerical gradient method of new frontends
• Higher order numerical derivatives
• Another numerical gradient function
• Numerical Hessian function
• A string, yielding the numerical gradient if evaluated
• Complex step derivatives