source: trunk/c++_tools/classifier/SVM.h @ 675

#ifndef _theplu_classifier_svm_ 
#define _theplu_classifier_svm_ 

// $Id$

/*
  Copyright (C) The authors contributing to this file.

  This file is part of the yat library, http://lev.thep.lu.se/trac/yat

  The yat library is free software; you can redistribute it and/or
  modify it under the terms of the GNU General Public License as
  published by the Free Software Foundation; either version 2 of the
  License, or (at your option) any later version.

  The yat library is distributed in the hope that it will be useful,
  but WITHOUT ANY WARRANTY; without even the implied warranty of
  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
  General Public License for more details.

  You should have received a copy of the GNU General Public License
  along with this program; if not, write to the Free Software
  Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA
  02111-1307, USA.
*/

#include "yat/classifier/KernelLookup.h"
#include "yat/classifier/SupervisedClassifier.h"
#include "yat/classifier/SVindex.h"
#include "yat/classifier/Target.h"
#include "yat/utility/vector.h"

#include <utility>
#include <vector>


namespace theplu {
namespace classifier {  

  class DataLookup2D;
  ///
  /// @brief Support Vector Machine 
  ///
  /// 
  ///
  /// Class for SVM using Keerthi's second modification of Platt's
  /// Sequential Minimal Optimization. The SVM uses all data given for
  /// training. If validation or testing is wanted this should be
  /// taken care of outside (in the kernel).
  ///   
  class SVM : public SupervisedClassifier
  {
  
  public:
    ///
    /// Constructor taking the kernel and the target vector as
    /// input. 
    ///
    /// @note if the @a target or @a kernel
    /// is destroyed the behaviour is undefined.
    ///
    SVM(const KernelLookup& kernel, const Target& target);

    ///
    /// Destructor
    ///
    virtual ~SVM();

    ///
    /// If DataLookup2D is not a KernelLookup a bad_cast exception is thrown.
    ///
    SupervisedClassifier* 
    make_classifier(const DataLookup2D&, const Target&) const;

    ///
    /// @return \f$ \alpha \f$
    ///
    inline const utility::vector& alpha(void) const { return alpha_; }

    ///
    /// The C-parameter is the balance term (see train()). A very
    /// large C means the training will be focused on getting samples
    /// correctly classified, with risk for overfitting and poor
    /// generalisation. A too small C will result in a training where
    /// misclassifications are not penalized. C is weighted with
    /// respect to the size, so \f$ n_+C_+ = n_-C_- \f$, meaning a
    /// misclassificaion of the smaller group is penalized
    /// harder. This balance is equivalent to the one occuring for
    /// regression with regularisation, or ANN-training with a
    /// weight-decay term. Default is C set to infinity.
    ///
    /// @returns mean of vector \f$ C_i \f$
    ///
    inline double C(void) const { return 1/C_inverse_; }

    ///
    /// Default is max_epochs set to 10,000,000.
    ///
    /// @return number of maximal epochs
    ///
    inline long int max_epochs(void) const {return max_epochs_;}
    
    /// 
    /// The output is calculated as \f$ o_i = \sum \alpha_j t_j K_{ij}
    /// + bias \f$, where \f$ t \f$ is the target.
    ///
    /// @return output
    /// 
    inline const theplu::utility::vector& output(void) const { return output_; }

    ///
    /// Generate prediction @a predict from @a input. The prediction
    /// is calculated as the output times the margin, i.e., geometric
    /// distance from decision hyperplane: \f$ \frac{ \sum \alpha_j
    /// t_j K_{ij} + bias}{w} \f$ The output has 2 rows. The first row
    /// is for binary target true, and the second is for binary target
    /// false. The second row is superfluous as it is the first row
    /// negated. It exist just to be aligned with multi-class
    /// SupervisedClassifiers. Each column in @a input and @a output
    /// corresponds to a sample to predict. Each row in @a input
    /// corresponds to a training sample, and more exactly row i in @a
    /// input should correspond to row i in KernelLookup that was used
    /// for training.
    ///
    void predict(const DataLookup2D& input, utility::matrix& predict) const;

    ///
    /// @return output times margin (i.e. geometric distance from
    /// decision hyperplane) from data @a input
    ///
    double predict(const DataLookup1D& input) const;

    ///
    /// @return output times margin from data @a input with
    /// corresponding @a weight
    ///
    double predict(const DataLookupWeighted1D& input) const;

    ///
    /// Function sets \f$ \alpha=0 \f$ and makes SVM untrained.
    ///
    inline void reset(void) 
    { trained_=false; alpha_=utility::vector(target_.size(),0); }

    ///
    /// @brief sets the C-Parameter
    ///
    void set_C(const double);

    /**
       Training the SVM following Platt's SMO, with Keerti's
       modifacation. Minimizing \f$ \frac{1}{2}\sum
       y_iy_j\alpha_i\alpha_j(K_{ij}+\frac{1}{C_i}\delta_{ij}) \f$ ,
       which corresponds to minimizing \f$ \sum w_i^2+\sum C_i\xi_i^2
       \f$.
    */
    bool train();

       
      
  private:
    ///
    /// Copy constructor. (not implemented)
    /// 
    SVM(const SVM&);
         
    ///
    /// Calculates bounds for alpha2
    ///
    void bounds(double&, double&) const;

    ///
    /// @brief calculates the bias term
    ///
    /// @return true if successful
    ///
    bool calculate_bias(void);

    ///
    /// Calculate margin that is inverse of w
    ///
    void calculate_margin(void);

    ///
    ///   Private function choosing which two elements that should be
    ///   updated. First checking for the biggest violation (output - target =
    ///   0) among support vectors (alpha!=0). If no violation was found check
    ///   sequentially among the other samples. If no violation there as
    ///   well training is completed
    ///
    ///  @return true if a pair of samples that violate the conditions
    ///  can be found
    ///
    bool choose(const theplu::utility::vector&);

    ///
    /// @return kernel modified with diagonal term (soft margin)
    ///
    inline double kernel_mod(const size_t i, const size_t j) const 
    { return i!=j ? (*kernel_)(i,j) : (*kernel_)(i,j) + C_inverse_; }
    
    /// @return 1 if i belong to binary target true else -1
    inline int target(size_t i) const { return target_.binary(i) ? 1 : -1; }

    utility::vector alpha_;
    double bias_;
    double C_inverse_;
    const KernelLookup* kernel_; 
    double margin_;
    unsigned long int max_epochs_;
    utility::vector output_;
    bool owner_;
    SVindex sample_;
    bool trained_;
    double tolerance_;

  };

}} // of namespace classifier and namespace theplu

#endif