ILModel Interface Reference

This interface provide functionalities of the machine learning model. More...

Inheritance diagram for ILModel:

Properties
LPVModelType	Type [get, set]
	The type of the model. Changing the type will reset the model to default.
Properties inherited from ILParameterized
BOOL	ParamBool (LString paramName) [get, set]
	The boolean parameter.
double	ParamF (LString paramName) [get, set]
	The floating double parameter.
int	ParamI (LString paramName) [get, set]
	The integer parameter.
LSize	ParamSize (LString paramName) [get, set]
	The size parameter.

Additional Inherited Members
Public Member Functions inherited from ILObject
ILObject *	Copy ()
LPVErrorCode	Load (LString filename)
void	Reset ()
LPVErrorCode	Save (LString filename)
BOOL	Valid ()

Detailed Description

This interface provide functionalities of the machine learning model.

You should firstly create the model with specified type via Create(), then adjust its parameters via ILParameterized interface.
Currently, we support following models:

Model	Usage and Parameters
KNN	K-Nearest Neighbors Parameter Name Usage Type Range Default "number of neighbors" Number of neighbors used for prediction Integer 1 ~ 64 5 "matching algorithm" The algorithm for the nearest searching Integer see LPVKNNMatchAlgo Brute-Force "similarity metric" The distance calculation method for measuring the similarity between samples Integer see LPVKNNDistance Euclidean "weight" How neighbors are weighted in prediction voting Integer see LPVKNNWeight By distance
SVM	Support Vector Machine Parameter Name Usage Type Range Default "svm type" The type of the SVM model Integer see LPVSVMType C-Support "cost" The regularization parameter C, the strength of the regularization is inversely proportional to C. Integer > 0 0.01 "complexity bound" The Nu parameter to control the number of support vectors and margin errors. It is and upper bound on the fraction of margin errors and a lower bound of the fraction of support vectors Double 0 ~ 1 0.5 "kernel" The kernel function Integer see LPVSVMKernel Linear "g" g parameter in the kernel function Double > 0, 0 for auto-estimated 0.1 "c" c parameter in the kernel function Double ≥ 0 0.1 "n" n parameter in the kernel function Double > 0 3

Example Code

C++

// create a HOG feature
ILModelPtr svm = LModel);
svm->Type = LPVModelType::LPVModelSVM;
// change its parameters
svm->PutParamI(L"svm type", LPVSVMType::LPVSVMTypeCSupport);
svm->PutParamF(L"cost", 512);
svm->PutParamI(L"kernel", LPVSVMKernel::LPVSVMKernelPolynomial);
svm->PutParamF(L"c", 0.4);
llc->AddModel(svm); // add it to the classifier

C#

// create a HOG feature
LModel svm = new LModel();
svm.Type = LPVModelType.LPVModelSVM;
// change its parameters
svm.ParamI["svm type"] = LPVSVMType.LPVSVMTypeCSupport;
svm.ParamF["cost"] = 512;
svm.ParamI["kernel"] = LPVSVMKernel.LPVSVMKernelPolynomial;
svm.ParamF["c"] = 0.4;
llc.AddModel(svm); // add it to the classifier

COM

// create a HOG feature
ILModelPtr svm = LModel::Create();
// or, ILModelPtr svm(__uuidof(LModel)), for absence of built-in Create functions (before 2.2.0)
svm->Type = LPVModelType::LPVModelSVM;
// change its parameters
svm->ParamI[L"svm type"] = LPVSVMType::LPVSVMTypeCSupport;
svm->ParamF[L"cost"] = 512;
svm->ParamI[L"kernel"] = LPVSVMKernel::LPVSVMKernelPolynomial;
svm->ParamF[L"c"] = 0.4;
llc->AddModel(svm); // add it to the classifier