Accord.MachineLearning.VectorMachines Namespace

Contains classes related to Support Vector Machines (SVMs). Contains linear machines, kernel machines, multi-class machines, SVM-DAGs (Directed Acyclic Graphs), multi-label classification and also offers support for the probabilistic output calibration of SVM outputs.

Classes

	Class	Description
	KernelSupportVectorMachine	Obsolete. Obsolete.
	MulticlassSupportVectorMachine	Obsolete. One-against-one Multi-class Kernel Support Vector Machine Classifier.
	MulticlassSupportVectorMachineTKernel	One-against-one Multi-class Kernel Support Vector Machine Classifier.
	MulticlassSupportVectorMachineTKernel, TInput	One-against-one Multi-class Kernel Support Vector Machine Classifier.
	MulticlassSupportVectorMachineTModel, TKernel, TInput	One-against-one Multi-class Kernel Support Vector Machine Classifier.
	MultilabelSupportVectorMachine	Obsolete. One-against-all Multi-label Kernel Support Vector Machine Classifier.
	MultilabelSupportVectorMachineTKernel	One-against-all Multi-label Kernel Support Vector Machine Classifier.
	MultilabelSupportVectorMachineTKernel, TInput	One-against-all Multi-label Kernel Support Vector Machine Classifier.
	MultilabelSupportVectorMachineTModel, TKernel, TInput	One-against-all Multi-label Kernel Support Vector Machine Classifier.
	SupportVectorMachine	Linear Support Vector Machine (SVM).
	SupportVectorMachineTKernel	Sparse Kernel Support Vector Machine (kSVM)
	SupportVectorMachineTKernel, TInput	Sparse Kernel Support Vector Machine (kSVM)

Interfaces

	Interface	Description
	ISupportVectorMachineTInput	Common interface for binary support vector machines.

Enumerations

	Enumeration	Description
	MulticlassComputeMethod	Decision strategies for Multi-class Support Vector Machines.
	MultilabelProbabilityMethod	Probability computation strategies for MultilabelSupportVectorMachine

Remarks

This namespace contains both standard SupportVectorMachines and the kernel extension given by SupportVectorMachineTKernels. For multiple classes or categories, the framework offers MulticlassSupportVectorMachineTKernels and MultilabelSupportVectorMachineTKernels. Multi-class machines can be used for cases where a single class should be picked up from a list of several class labels, and the multi-label machine for cases where multiple class labels might be detected for a single input vector. The multi-class machines also support two types of classification: the faster decision based on Decision Directed Acyclic Graphs, and the more traditional based on a Voting scheme.

Learning can be achieved using the standard SequentialMinimalOptimizationTKernel (SMO) algorithm. However, the framework can also learn Least Squares SVMs (LS-SVMs) using LeastSquaresLearning, and even calibrate SVMs to produce probabilistic outputs using ProbabilisticOutputCalibration. A huge variety of kernels functions is available in the statistics namespace, and new kernels can be created easily using the IKernel interface.

The namespace class diagram is shown below.

Please note that class diagrams for each of the inner namespaces are also available within their own documentation pages.

Reference

Accord.MachineLearning.VectorMachines.Learning

Accord.Statistics.Kernels