stdfmin

stdfmin is a generic term used in several numerical optimization libraries for a routine that minimizes a standard-form objective function. The exact interface and capabilities of stdfmin vary by implementation, but the common purpose is to find a vector x* in R^n that minimizes f(x). Typical usage involves providing a function handle or callable that computes f(x), an initial guess x0, and optional parameters such as tolerances, maximum iterations, and information about gradients or Hessians if available.

Output usually includes the estimated minimizer x*, the minimum value f(x*), and an exit status or diagnostic information (e.g., success flag, number of iterations, gradient norm).

Algorithmically, stdfmin implementations may use gradient-based methods (quasi-Newton like BFGS/L-BFGS, conjugate gradient) when derivatives are supplied, or derivative-free methods (Nelder-Mead, Powell) when not. Some variants combine line search or trust-region strategies to improve robustness and to handle nonlinearity or ill-conditioning. Many versions also offer simple bound constraints or variable scaling.

In practice, stdfmin is used for parameter estimation, model fitting, and other optimization problems where the objective is a standard form function to be minimized. Its behavior is problem-dependent and no single standard exists, so users should consult library-specific documentation to understand supported features, input formats, and default settings.

See also: optimization, unconstrained minimization, fmin, Nelder-Mead, BFGS.