template<class A> class NMSimplex

Need to be derived to evaluate the apropiate fuction in each step

Public Methods

NMSimplex()

default constructor (dim=dimension of the problem (default=2) )

virtual ~NMSimplex()

destructor

void SetDimensions(int dim)

This sets the dimensions of all vectorsNS, simplex.

void SetAlpha(double newAlpha)

These functions allow the user to set values of the reflection, contraction, expansion, and shrinking coefficients

void SetBeta(double newBeta)

void SetGamma(double newGamma)

void SetSigma(double newSigma)

void SetMaxCalls(long mcalls)

void ExploratoryMoves()

Use Nelder Mead to find function minimum

void ReplaceSimplexPoint(int index, const VectorND& newPoint)

Replaces simplex point indexed at index with newPoint

void CalculateFunctionValue(int index)

Finds the f(x) value for the simplex point indexed at index and replaces the proper value in simplexValues

bool Stop()

Returns true if the stopping criteria have been satisfied

void InitRegularTriangularSimplex(const VectorND basePoint, const double edgeLength, A &a)

Simplex-altering functions: Initialization

void InitFixedLengthRightSimplex(const VectorND basePoint, const double edgeLength, A &a)

NOTE: basePoint is assumed to be of proper dimension

void InitVariableLengthRightSimplex(const VectorND basePoint, const double* edgeLengths, A &a)

NOTE: basePoint is assumed to be of proper dimension

void InitGeneralSimplex(const MatrixNM* plex)

NOTE: basePoint and edgeLengths are assumed to be of proper dimension

int GetFunctionCalls() const

number of objective function evaluations

int GetVarNo() const

returns the dimension of the problem

int GetTolHit() const

returns toleranceHit

void printSimplex() const

prints out the simplex points by row, their corresponding f(x) values, and the number of function calls thus far

VectorND GetMinPoint() const

simplex point which generates the best objective function value found thus far

double GetMinVal() const

best objective function value found thus far

Protected Fields

A* x

-----------------------------------------------------------

Protected Methods

virtual void fcnCall(VectorND v, double& value, int& flag)

------------ Modify this call of objective function for template design

Private Fields

int dimensions

MatrixNM* simplex

the number of dimensions (the dimension of the problem)

double* simplexValues

the current simplex

double alpha

their corresponding f(x) values

double beta

reflection coefficient

double gamma

contraction coefficient

double sigma

expansion coefficient

int minIndex

shrinking coefficient

int maxIndex

index of point generating min f(x)

VectorND* centroid

index of point generating max f(x)

VectorND* reflectionPt

the current centroid

double reflectionPtValue

the reflection point

VectorND* expansionPt

the value of f(reflectionPt)

double expansionPtValue

the expansion point

VectorND* contractionPt

the value of f(expansionPt)

double contractionPtValue

the contraction point

double maxPrimePtValue

the value of f(contractionPt)

long functionCalls

min(f(maxIndexPoint),reflectionPtValue)

int toleranceHit

tally of number of function calls

int maxPrimePtId

1 if stop due to tolerance, 0 if funcCalls, -1 if not stopped

long MaxCalls

The maxCalls variable is not a hard limit on the number of function evaluations that may be performed by the search.

Private Methods

void FindMinMaxIndices()

sets maxIndex to the simplex index of the point which generates the highest value of f(x)

int SecondHighestPtIndex()

returns simplex index of the point which generates the second highest value of f(x)

void FindCentroid()

finds the centroid of the simplex

void FindReflectionPt()

finds the reflection point and sets its f(x) value

void FindExpansionPt()

finds the expansion point and sets its f(x) value

void FindContractionPt()

finds the contraction point and sets its f(x) value

void ShrinkSimplex()

this function goes through the simplex and reduces the lengths of the edges adjacent to the best vertex

Documentation

Need to be derived to evaluate the apropiate fuction in each step

NMSimplex()

default constructor (dim=dimension of the problem (default=2) )

virtual ~NMSimplex()

destructor

void SetDimensions(int dim)

This sets the dimensions of all vectorsNS, simplex. Call always after Construction if dimensions!=2.

void SetAlpha(double newAlpha)

These functions allow the user to set values of the reflection, contraction, expansion, and shrinking coefficients

void SetBeta(double newBeta)

void SetGamma(double newGamma)

void SetSigma(double newSigma)

void SetMaxCalls(long mcalls)

void ExploratoryMoves()

Use Nelder Mead to find function minimum

void ReplaceSimplexPoint(int index, const VectorND& newPoint)

Replaces simplex point indexed at index with newPoint

void CalculateFunctionValue(int index)

Finds the f(x) value for the simplex point indexed at index and replaces the proper value in simplexValues

bool Stop()

Returns true if the stopping criteria have been satisfied

void InitRegularTriangularSimplex(const VectorND basePoint, const double edgeLength, A &a)

Simplex-altering functions: Initialization

void InitFixedLengthRightSimplex(const VectorND basePoint, const double edgeLength, A &a)

NOTE: basePoint is assumed to be of proper dimension

void InitVariableLengthRightSimplex(const VectorND basePoint, const double* edgeLengths, A &a)

NOTE: basePoint is assumed to be of proper dimension

void InitGeneralSimplex(const MatrixNM* plex)

NOTE: basePoint and edgeLengths are assumed to be of proper dimension

int GetFunctionCalls() const

number of objective function evaluations

int GetVarNo() const

returns the dimension of the problem

int GetTolHit() const

returns toleranceHit

void printSimplex() const

prints out the simplex points by row, their corresponding f(x) values, and the number of function calls thus far

VectorND GetMinPoint() const

simplex point which generates the best objective function value found thus far

double GetMinVal() const

best objective function value found thus far

virtual void fcnCall(VectorND v, double& value, int& flag)

------------ Modify this call of objective function for template design

A* x

-----------------------------------------------------------

void FindMinMaxIndices()

sets maxIndex to the simplex index of the point which generates the highest value of f(x)

int SecondHighestPtIndex()

returns simplex index of the point which generates the second highest value of f(x)

void FindCentroid()

finds the centroid of the simplex

void FindReflectionPt()

finds the reflection point and sets its f(x) value

void FindExpansionPt()

finds the expansion point and sets its f(x) value

void FindContractionPt()

finds the contraction point and sets its f(x) value

void ShrinkSimplex()

this function goes through the simplex and reduces the lengths of the edges adjacent to the best vertex

int dimensions

MatrixNM* simplex

the number of dimensions (the dimension of the problem)

double* simplexValues

the current simplex

double alpha

their corresponding f(x) values

double beta

reflection coefficient

double gamma

contraction coefficient

double sigma

expansion coefficient

int minIndex

shrinking coefficient

int maxIndex

index of point generating min f(x)

VectorND* centroid

index of point generating max f(x)

VectorND* reflectionPt

the current centroid

double reflectionPtValue

the reflection point

VectorND* expansionPt

the value of f(reflectionPt)

double expansionPtValue

the expansion point

VectorND* contractionPt

the value of f(expansionPt)

double contractionPtValue

the contraction point

double maxPrimePtValue

the value of f(contractionPt)

long functionCalls

min(f(maxIndexPoint),reflectionPtValue)

int toleranceHit

tally of number of function calls

int maxPrimePtId

1 if stop due to tolerance, 0 if funcCalls, -1 if not stopped

long MaxCalls

The maxCalls variable is not a hard limit on the number of function evaluations that may be performed by the search. The search should always halt at a point from which none of the trial steps can find improvement to be consistent with the formal theory regarding pattern searches.

This class has no child classes.

Alphabetic index HTML hierarchy of classes or Java