Svm.md

Support-Vector Machine CSvm

• Support-Vector Machine CSvm
  • Training settings
  • Model
  • Sample

Support-vector machine translates the input data into vectors in a high-dimensional space and searches for a maximum-margin dividing hyperplane.

In NeoML library this method is implemented by the CSvm class. It exposes a Train method for creating a model for binary classification.

Training settings

The parameters are represented by a CSvm::CParams structure.

• KernelType — the type of the kernel function
• ErrorWeight — the error weight relative to the regularization function
• MaxIterations — the maximum number of algorithm iterations
• Degree — the degree for a gaussian kernel
• Gamma — the kernel coefficient (for KT_Poly, KT_RBF, KT_Sigmoid)
• Coeff0 — the kernel free term (for KT_Poly, KT_Sigmoid)
• Tolerance — the algorithm precision, the stop criterion
• ThreadCount — the number of processing threads to be used while training
• MulticlassMode - the approach used in multiclass task: OneVsAll (default) or OneVsOne

Model

The trained model implements the ILinearBinaryModel interface if a KT_Linear kernel is used; or MulticlassMode model if number of classes > 2; otherwise, it implements the ISvmBinaryModel interface.

// SVM binary classifier interface
class ISvmBinaryModel : public IModel {
public:
	virtual ~ISvmBinaryModel();

	// Get the kernel
	virtual CSvmKernel::TKernelType GetKernelType() const = 0;

	// Get the support vectors
	virtual CSparseFloatMatrix GetVectors() const = 0;

	// Get the support vector coefficients
	virtual const CArray<double>& GetAlphas() const = 0;

	// Get the free term
	virtual double GetFreeTerm() const = 0;

	// Serialize the model
	virtual void Serialize( CArchive& ) = 0;
};

Sample

Here is a simple example of training a support-vector classification model. The input data is represented by an object implementing the IProblem interface.

CPtr<Model> buildModel( IProblem* data )
{
	CSvm::CParams params( CSvmKernel::KT_RBF );
	CSvm builder( params );
	return builder.Train( *data );
}