Logo ROOT   6.08/07
Reference Guide
Functions
ROOT::Math::BrentMethods Namespace Reference

Functions

double MinimBrent (const IGenFunction *f, int type, double &xmin, double &xmax, double xmiddle, double fy, bool &ok, int &niter, double epsabs=1.E-8, double epsrel=1.E-10, int maxiter=100)
 Finds a minimum of a function, if the function is unimodal between xmin and xmax This method uses a combination of golden section search and parabolic interpolation Details about convergence and properties of this algorithm can be found in the book by R.P.Brent "Algorithms for Minimization Without Derivatives" or in the "Numerical Recipes", chapter 10.2 convergence is reached using tolerance = 2 *( epsrel * abs(x) + epsabs) More...
 
double MinimStep (const IGenFunction *f, int type, double &xmin, double &xmax, double fy, int npx=100, bool useLog=false)
 Grid search implementation, used to bracket the minimum and later use Brent's method with the bracketed interval The step of the search is set to (xmax-xmin)/fNpx type: 0-returns MinimumX 1-returns Minimum 2-returns MaximumX 3-returns Maximum 4-returns X corresponding to fy. More...
 

Function Documentation

◆ MinimBrent()

double ROOT::Math::BrentMethods::MinimBrent ( const IGenFunction f,
int  type,
double &  xmin,
double &  xmax,
double  xmiddle,
double  fy,
bool &  ok,
int &  niter,
double  epsabs = 1.E-8,
double  epsrel = 1.E-10,
int  maxiter = 100 
)

Finds a minimum of a function, if the function is unimodal between xmin and xmax This method uses a combination of golden section search and parabolic interpolation Details about convergence and properties of this algorithm can be found in the book by R.P.Brent "Algorithms for Minimization Without Derivatives" or in the "Numerical Recipes", chapter 10.2 convergence is reached using tolerance = 2 *( epsrel * abs(x) + epsabs)

type: 0-returns MinimumX 1-returns Minimum 2-returns MaximumX 3-returns Maximum 4-returns X corresponding to fy

if ok=true the method has converged. Maxiter returns the actual number of iteration performed

Definition at line 85 of file BrentMethods.cxx.

◆ MinimStep()

double ROOT::Math::BrentMethods::MinimStep ( const IGenFunction f,
int  type,
double &  xmin,
double &  xmax,
double  fy,
int  npx = 100,
bool  useLog = false 
)

Grid search implementation, used to bracket the minimum and later use Brent's method with the bracketed interval The step of the search is set to (xmax-xmin)/fNpx type: 0-returns MinimumX 1-returns Minimum 2-returns MaximumX 3-returns Maximum 4-returns X corresponding to fy.

Definition at line 29 of file BrentMethods.cxx.