Rivet::MendelMin Class Reference

A genetic algorithm functional minimizer. More...

#include <MendelMin.hh>

Public Types
using	Params = std::valarray<double>
	Typedef for a valaray of parameters to the function to be minimised.
using	FuncT = std::function<double(const Params&, const Params&)>
	Typedef for the function to be minimised.
using	FuncNoFixedT = std::function<double(const Params&)>
	Typedef for the function to be minimised.

Public Member Functions
	MendelMin (const FuncT &fin, unsigned int ndim, const Params &fixpar, unsigned int npop=20, double margin=0.1)
	MendelMin (const FuncNoFixedT &fin, unsigned int ndim, unsigned int npop=20, unsigned int ngen=20, double margin=0.1)
void	guess (const Params &p)
double	evolve (unsigned int nGen)
Params	fittest () const
	Return the fittest parameter point found.
double	fit () const
	Return the fittest value found.
double	rnd () const
	Simple wrapper around the random number generator.
Params	rndParams () const
	Return a random parameter point in the unit hypercube.
void	limit01 (Params &p) const
	Limit a parameter point to inside the unit hypercube.
void	move (Params &bad, const Params &better) const
double	f (const Params &p) const
	Simple wrapper around the function to be minimised.
std::pair< double, double >	minmax ()

Public Attributes
bool	showTrace
	Set true to get a verbose record of the evolution.

Detailed Description

A genetic algorithm functional minimizer.

MendelMin implements a home brewed genetic algorithm for finding the minimum of a function defined on a unit hypercube returning a non-negative real number (eg. a Chi-squared value).

Constructor & Destructor Documentation

MendelMin() [1/2]

Rivet::MendelMin::MendelMin ( const FuncT & fin, unsigned int ndim, const Params & fixpar, unsigned int npop = 20, double margin = 0.1 )

inline

Constructor with fixed parameters

Mandatory arguments: the function, fin, to be minimised; the dimension, ndim, of the unit hypercube for which fin is defined; a set of fixed parameters not to be optimised.

Optional arguments are: the number, npop, of individuals in the population; and margin which determines how much randomness is involved when an individual is evolved twowards the fittest individual.

References showTrace.

Referenced by MendelMin().

MendelMin() [2/2]

Rivet::MendelMin::MendelMin ( const FuncNoFixedT & fin, unsigned int ndim, unsigned int npop = 20, unsigned int ngen = 20, double margin = 0.1 )

inline

Constructor without fixed parameters

Mandatory arguments: the function, fin, to be minimised; the dimension, ndim, of the unit hypercube for which fin is defined.

Optional arguments are: the number, npop, of individuals in the population; and margin which determines how much randomness is involved when an individual is evolved twowards the fittest individual.

References MendelMin().

Member Function Documentation

evolve()

double Rivet::MendelMin::evolve ( unsigned int nGen )

inline

Evolve the population a given number of generations and return the best fit value.

References Rivet::minmax(), move(), rnd(), and showTrace.

guess()

void Rivet::MendelMin::guess ( const Params & p )

inline

Supply a best guess for the fittest parameter point to help things along.

References f(), limit01(), and Rivet::Kin::p().

minmax()

std::pair< double, double > Rivet::MendelMin::minmax ( )

inline

Calculate the fitness values of all individuals and put the fittest one first.

Returns

the best and worst fitness values.

References f(), and rndParams().

move()

void Rivet::MendelMin::move ( Params & bad, const Params & better ) const

inline

Move a bad parameter point towards a better one. The new point is picked randomly within the generalized hypercube where bad and better are at diagonally opposite corners, enlarged by a fraction _margin.

References limit01(), and rndParams().

Referenced by evolve().

---

The documentation for this class was generated from the following file:

• /builds/hepcedar/rivet/include/Rivet/Tools/MendelMin.hh