<< opt opt opt_ic >>

CCA (Computational Convex Analysis) >> opt > opt_co

opt_co

Convex operators via optimization (OPT), Convex hull

Calling Sequence

[y, argmin] = opt_me(f, df, domf, x, output)

Parameters

f

Univariate function, continuous but (potentially) nonsmooth.

df

Univariate function: derivative of f.

domf

Domain of the function f (set of points where the function is finite). Domains are stored as intervals I=[lb,ub] with ub<lb storing an empty set.

s

Vector of points at which to evaluate the convex hull.

output

0: no output, 1: progress bar.

y

The values of the convex hull evaluated at each x.

argmin

A point at which the convex hull attains its minimum, for each x.

Description

CURRENTLY BROKEN: NEED A NONCONVEX NONSMOOTH SOLVER (Neldermead in Scilab 5.2.0: fminsearch with infinite penalization or neldermead with Box complex algorithm). BUT optim WON'T WORK SINCE 'nd' IS A BUNDLE METHOD (only works for convex functions).

Computes the convex hull of a (potentially) nonconvex function f, evaluated on vector s, giving the values y such that

where x1,x2,lambda are vectors of the same dimension as x.

Authors

<< opt opt opt_ic >>