TIEVisionMath.optimizeWith_NelderMeadSimplexMethod
Declaration
function optimizeWith_NelderMeadSimplexMethod(func: TIEVisionOptimizerFunction; funcUserData: pointer; solutions: TIEVisionVectorDouble; out optimalValue: double; length: double = 1.0; timeout: int32_t = 1000; eps: double = 1.0e-20): bool32; safecall;Description
Optimizes (minimizes) the specified function using Nelder Mead Simplex method.Returns true on success.
| Parameter | Description |
|---|---|
| func | Function to optimize (minimize) |
| funcUserData | A user pointer sent to function to optimize |
| solutions | Initial veriables and output minimizing variables |
| optimalValue | Optimal value |
| length | Initial length of the simplex |
| timeout | Maximum number of iterations |
| eps | Small real number to test convergence |
Demos
 | Demos\IEVision\GetFaces\GetFaces.dpr |
| | Demos\IEVision\GetWithClassifier\GetMisc.dpr |
| | Demos\IEVision\NumberPlateBlurring\GetPlate.dpr |
| | Demos\IEVision\FaceDetection\FaceDetection.dpr |
| | Demos\IEVision\FaceDetection_LowLevel\FaceDetection_LowLevel.dpr |
| | Demos\IEVision\TrackObjects\TrackObjects.dpr |
| | Demos\IEVision\TrackObjects_LowLevel\TrackObjects_LowLevel.dpr |
Example
// finds values (x0=vars[0] and x1=vars[1]) that minimze (x0-1)*(x0-1) + (x1-6)*(x1-6)function optFunc(userData: pointer; vars: TIEVisionVectorDouble): double; stdcall;
var
x0, x1: double;
begin
x0 := vars.getDouble(0);
x1 := vars.getDouble(1);
result := (x0-1)*(x0-1) + (x1-6)*(x1-6);
end;
....
// vector of input and output variables
vec := IEVisionLib.createVectorDouble();
// initial values
vec.push_back(-1.2);
vec.push_back(1.0);
// find minimizing values
if not IEVisionLib.createMath().optimizeWith_NelderMeadSimplexMethod(@optFunc, nil, vec, optValue, 1.0, 1000, 1e-20) then
begin
x0 := vec.getDouble(0); // should be "1"
x1 := vec.getDouble(1); // should be "6"
// ...and optValue shoud be "0"
end;
....