algorithms.optimize

Module: algorithms.optimize

nipy.algorithms.optimize.fmin_steepest(f, x0, fprime=None, xtol=0.0001, ftol=0.0001, maxiter=None, epsilon=1.4901161193847656e-08, callback=None, disp=True)

Minimize a function using a steepest gradient descent algorithm. This complements the collection of minimization routines provided in scipy.optimize. Steepest gradient iterations are cheaper than in the conjugate gradient or Newton methods, hence convergence may sometimes turn out faster algthough more iterations are typically needed.

Parameters:

f : callable

Function to be minimized

x0 : array

Starting point

fprime : callable

Function that computes the gradient of f

xtol : float

Relative tolerance on step sizes in line searches

ftol : float

Relative tolerance on function variations

maxiter : int

Maximum number of iterations

epsilon : float or ndarray

If fprime is approximated, use this value for the step

size (can be scalar or vector).

callback : callable

Optional function called after each iteration is complete

disp : bool

Print convergence message if True

Returns:

x : array

Gradient descent fix point, local minimizer of f