recursive_least_squares.hpp

Go to the documentation of this file.

#ifndef PCRLS_HPP
#define PCRLS_HPP
 
#include <Eigen/Dense>
#include <mx/improc/eigenCube.hpp>
#include <mx/improc/eigenImage.hpp>
using namespace mx::improc;
#include "utils.hpp"
 
namespace DDSPC
{
 
/*
   This is just a generic Recursive Least Squares implementation.
   The auto-regressive model will be implemented in a different class.
   This allows for better reuse of the Recursive Least Squares.
*/
 
class RecursiveLeastSquares{
    
   private:
 
   public:
      realT _gamma;
        realT _inverse_gamma;
      realT _initial_covariance;
      int _num_features;
        int _num_predictors;
 
        Matrix K;           // The gain matrix
      Matrix err;         // The a-priori prediction error
      Matrix xtP;         // A convenience variable
        
      
      Matrix prediction_matrix;         // Prediction matrix
      Matrix inverse_covariance;       // Inverse covariance
 
        Matrix prediction_output;
      
      RecursiveLeastSquares(int num_predictors, int num_features, realT forgetting_factor, realT inverse_covariance);
      ~RecursiveLeastSquares();
 
        // This interface might need to change
      void update(eigenImage<realT> *x, eigenImage<realT> *y);
      void update(Matrix *x, Matrix *y);
        
        //Matrix* predict(eigenImage<realT> *x);
      void reset();
      void save_state(std::string filaname);
};
 
}
 
#endif