Metrics.hpp

Go to the documentation of this file.

//
// Created by dodo on 31/10/17.
//

#ifndef MACHINE_LEARNING_METRICS_HPP
#define MACHINE_LEARNING_METRICS_HPP

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

class Metrics {

 public:

  static MatrixD minkowski(MatrixD m, double p, bool root = true) {
    MatrixD distances = MatrixD::zeros(m.nRows(), m.nRows());

    for (size_t i = 0; i < m.nRows(); i++) {
      for (size_t j = i + 1; j < m.nRows(); j++) {
        for (size_t k = 0; k < m.nCols(); k++) {
          distances(i, j) += pow(abs(m(i, k) - m(j, k)), p);
        }
        if (root)
          distances(i, j) = pow(distances(i, j), 1 / p);
      }
    }

    return distances;
  }

  static MatrixD chebyshev(MatrixD a) {
    MatrixD distances = MatrixD::zeros(a.nRows(), a.nRows());

    for (size_t i = 0; i < a.nRows(); i++) {
      for (size_t j = i + 1; j < a.nRows(); j++) {
        for (size_t k = 0; k < a.nCols(); k++) {
          double x1 = a(i, k), x2 = a(j, k);
          double term = abs(x1 - x2);
          distances(i, j) = distances(j, i) = max(distances(i, j), term);
        }
      }
    }

    return distances;
  }

  static MatrixD euclidean(MatrixD m, bool root = true) {
    return minkowski(m, 2, root);
  }

  static MatrixD manhattan(MatrixD m, bool root = true) {
    return minkowski(m, 1, root);
  }

  static MatrixD minkowski(MatrixD a, MatrixD b, double p, bool root = true) {
    if (a.nCols() != b.nCols())
      throw runtime_error("Matrices have different number of dimensions");

    MatrixD distances = MatrixD::zeros(a.nRows(), b.nRows());

    for (size_t i = 0; i < a.nRows(); i++) {
      for (size_t j = 0; j < b.nRows(); j++) {
        for (size_t k = 0; k < a.nCols(); k++) {
          double x1 = a(i, k), x2 = b(j, k);
          double term = pow(abs(x1 - x2), p);
          distances(i, j) += term;
        }
        if (root)

          distances(i, j) = pow(distances(i, j), 1 / p);
      }
    }

    return distances;
  }

  static MatrixD chebyshev(MatrixD a, MatrixD b) {
    if (a.nCols() != b.nCols())
      throw runtime_error("Matrices have different number of dimensions");

    MatrixD distances = MatrixD::zeros(a.nRows(), b.nRows());

    for (size_t i = 0; i < a.nRows(); i++) {
      for (size_t j = 0; j < b.nRows(); j++) {
        for (size_t k = 0; k < a.nCols(); k++) {
          double x1 = a(i, k), x2 = b(j, k);
          double term = abs(x1 - x2);
          distances(i, j) = max(distances(i, j), term);
        }
      }
    }

    return distances;
  }

  static MatrixD euclidean(MatrixD a, MatrixD b, bool root = true) {
    return minkowski(a, b, 2, root);
  }

  static MatrixD manhattan(MatrixD a, MatrixD b, bool root = true) {
    return minkowski(a, b, 1, root);
  }
};

#endif //MACHINE_LEARNING_METRICS_HPP