source: trunk/src/SVM.h @ 37

// $Id: SVM.h 37 2004-02-13 15:46:31Z peter $

#ifndef CS_CPP_TOOLS_SVM_H
#define CS_CPP_TOOLS_SVM_H

// C++ tools include
/////////////////////
#include "vector.h"
#include "matrix.h"


// Standard C++ includes
////////////////////////


namespace thep_cpp_tools 
{
  
  class SVM
  {
  
  public:
    /**
       Constructor taking the kernel matrix and the target vector as input
    */
    SVM(const thep_gsl_api::matrix&, const thep_gsl_api::vector&);
           
    /**
       Training the SVM using the SMO algorithm
    */
    void train();

    /**
       Function will return \f$\alpha\f$
    */
    inline thep_gsl_api::vector get_alpha() const;

    /**
       Function will return the output from SVM
    */
    inline thep_gsl_api::vector get_output() const;

   
      
  private:
    bool trained_;
    thep_gsl_api::matrix kernel_;
    thep_gsl_api::vector target_;
    thep_gsl_api::vector alpha_;
    double bias_;

    /**
       Private function that determines when to stop the training.
       The test is done in two steps. First, we check that the
       functional margin is at least 2 - epsilon. Second, we check
       that the gap between the primal and the dual object is less
       than epsilon.

    */
    bool SVM::stop(const thep_gsl_api::vector& target_,
       const thep_gsl_api::matrix& kernel_,
       const thep_gsl_api::vector& alpha_);
  };
  
    // class SVM
  thep_gsl_api::vector SVM::get_alpha() const  
  {
    return alpha_;
  }

  thep_gsl_api::vector SVM::get_output() const  
  {
    thep_gsl_api::vector bias(target_.size(), false, false);
    bias.set_all(bias_);
    return kernel_ * alpha_.mul_elements(target_) + bias;
    
  }
}; // namespace thep_c++_tools

#endif