Name: minimize-golden-section-1d
Owner: scijs
Description: Minimize a function of a single variable using golden section search
Created: 2016-05-09 06:22:53.0
Updated: 2018-03-20 00:44:16.0
Pushed: 2017-06-23 01:05:54.0
Homepage: null
Size: 10
Language: JavaScript
GitHub Committers
User | Most Recent Commit | # Commits |
---|
Other Committers
User | Most Recent Commit | # Commits |
---|
Minimize a function of a single variable using golden section search
Returns the argument that minimizes a function of a single variable using golden section search, a search algorithm similar to bisection search except that the size of the intervals is chosen so that function evaluations can be reused more effectively. The golden section works fine with discontinuities, asymptotes, or oscillations that would throw off derivative-based methods but is somewhat slower to converge, likely requiring more function evaluations.
m install minimize-golden-section-1d
Returns the argument that minimizes the function:
minimize = require('minimize-golden-section-1d');
mize(Math.cos)
> 3.1415926622945616
mize(Math.cos, {lowerBound: 0, upperBound: 1});
> 1
mize(Math.cos, {guess: -3});
> -3.1415926432597825
require('minimize-golden-section-1d')(f[, options])
Given function f
of one Number
-valued variable, computes a local minimum. On successful completion, returns the value of the argument that minimizes f
(note that this may either be a local or global minimum in the provided range). If the algorithm fails (e.g. NaN encountered or unconstrained divergence of the argument to Infinity
), returns NaN
. If tolerance is not met, returns best answer.
If bounds are provided, this module proceeds immediately with golden section search. If upper, lower or both bounds are not provided, the algorithm will make use of an initial guess (a provided guess or just one of the bounds if that's all it has) and expand the search range until a minimum is bracketed.
tolerance [=1e-8]
: Convergence tolerance. Algorithm continues until search interval is smaller than tolerance
.lowerBound [=-Infinity]
: Lower bound of the search intervalupperBound [=Infinity]
: Upper bound of the search intervalmaxIter [=100]
: Maximum number of iterations for either bracketing or search phase.guess [=0]
: initial guess for unconstrained minimization. Unused unless both upperBound
and lowerBound
are not provided.initialIncrement [=1]
: Initial interval by which to expand the search interval for unconstrained minimization. Unused unless both upperBound
and lowerBound
are not provided.© 2015 Scijs. MIT License.
Ricky Reusser