LDA.hpp

Go to the documentation of this file.

#ifndef MACHINE_LEARNING_LDA_HPP
#define MACHINE_LEARNING_LDA_HPP

#include <utility>

#include "../include/matrix/Matrix.hpp"

using namespace std;

class LDA {
 private:
  MatrixD X, y, eigenvalues, eigenvectors, transformedData;
 public:
  LDA(MatrixD data, MatrixD classes) : X(std::move(data)), y(std::move(classes)) {
    if (data.nRows() != classes.nRows())
      throw invalid_argument("data and classes must have the same number of rows");
    if (classes.nCols() != 1)
      throw invalid_argument("classes must me a column vector");
  }

  void fit() {
    MatrixD Sw = X.WithinClassScatter(y);
    MatrixD Sb = X.BetweenClassScatter(y);

    auto eigen = (Sw.inverse() * Sb).eigen();

    eigenvalues = eigen.first;
    eigenvectors = eigen.second;

    transformedData = (eigenvectors.transpose() * X.transpose()).transpose();
  }

  MatrixD transform() {
    return transformedData;
  }
};

#endif //MACHINE_LEARNING_LDA_HPP