AUTH's THMMY "Parallel and distributed systems" course assignments.
Вы не можете выбрать более 25 тем Темы должны начинаться с буквы или цифры, могут содержать дефисы(-) и должны содержать не более 35 символов.
 
 
 
 
 

120 строки
3.5 KiB

  1. /**
  2. * \file v0.hpp
  3. * \brief
  4. *
  5. * \author
  6. * Christos Choutouridis AEM:8997
  7. * <cchoutou@ece.auth.gr>
  8. */
  9. #ifndef V0_HPP_
  10. #define V0_HPP_
  11. #include <cblas.h>
  12. #include <cmath>
  13. #include <vector>
  14. #include <algorithm>
  15. #include <matrix.hpp>
  16. #include <config.h>
  17. namespace v0 {
  18. /*!
  19. * Function to compute squared Euclidean distances
  20. *
  21. * \fn void pdist2(const double*, const double*, double*, int, int, int)
  22. * \param X m x d matrix (Column major)
  23. * \param Y n x d matrix (Column major)
  24. * \param D2 m x n matrix to store distances (Column major)
  25. * \param m number of rows in X
  26. * \param n number of rows in Y
  27. * \param d number of columns in both X and Y
  28. */
  29. template<typename DataType>
  30. void pdist2(const mtx::Matrix<DataType>& X, const mtx::Matrix<DataType>& Y, mtx::Matrix<DataType>& D2) {
  31. int M = X.rows();
  32. int N = Y.rows();
  33. int d = X.columns();
  34. // Compute the squared norms of each row in X and Y
  35. std::vector<DataType> X_norms(M), Y_norms(N);
  36. for (int i = 0; i < M ; ++i) {
  37. X_norms[i] = cblas_ddot(d, X.data() + i * d, 1, X.data() + i * d, 1);
  38. }
  39. for (int j = 0; j < N ; ++j) {
  40. Y_norms[j] = cblas_ddot(d, Y.data() + j * d, 1, Y.data() + j * d, 1);
  41. }
  42. // Compute -2 * X * Y'
  43. cblas_dgemm(CblasRowMajor, CblasNoTrans, CblasTrans, M, N, d, -2.0, X.data(), d, Y.data(), d, 0.0, D2.data(), N);
  44. // Step 3: Add the squared norms to each entry in D2
  45. for (int i = 0; i < M ; ++i) {
  46. for (int j = 0; j < N; ++j) {
  47. D2.set(D2.get(i, j) + X_norms[i] + Y_norms[j], i, j);
  48. //D2.set(std::max(D2.get(i, j), 0.0), i, j); // Ensure non-negative
  49. D2.set(std::sqrt(D2.get(i, j)), i, j); // Take the square root of each
  50. }
  51. }
  52. }
  53. template<typename DataType, typename IndexType>
  54. void quickselect(std::vector<std::pair<DataType, IndexType>>& vec, int k) {
  55. std::nth_element(
  56. vec.begin(),
  57. vec.begin() + k,
  58. vec.end(),
  59. [](const std::pair<DataType, IndexType>& a, const std::pair<DataType, IndexType>& b) {
  60. return a.first < b.first;
  61. });
  62. vec.resize(k); // Keep only the k smallest elements
  63. }
  64. /*!
  65. * \param C Is a MxD matrix (Corpus)
  66. * \param Q Is a NxD matrix (Query)
  67. * \param k The number of nearest neighbors needed
  68. * \param idx Is the Nxk matrix with the k indexes of the C points, that are
  69. * neighbors of the nth point of Q
  70. * \param dst Is the Nxk matrix with the k distances to the C points of the nth
  71. * point of Q
  72. */
  73. template<typename DataType, typename IndexType>
  74. void knnsearch(const mtx::Matrix<DataType>& C, const mtx::Matrix<DataType>& Q, int k,
  75. mtx::Matrix<IndexType>& idx,
  76. mtx::Matrix<DataType>& dst) {
  77. int M = C.rows();
  78. int N = Q.rows();
  79. mtx::Matrix<DataType> D(M, N);
  80. pdist2(C, Q, D);
  81. idx.resize(N, k);
  82. dst.resize(N, k);
  83. for (int j = 0; j < N; ++j) {
  84. // Create a vector of pairs (distance, index) for the j-th query
  85. std::vector<std::pair<DataType, IndexType>> dst_idx(M);
  86. for (int i = 0; i < M; ++i) {
  87. dst_idx[i] = {D.data()[i * N + j], i};
  88. }
  89. // Find the k smallest distances using quickSelectKSmallest
  90. quickselect(dst_idx, k);
  91. // Sort the k smallest results by distance for consistency
  92. std::sort(dst_idx.begin(), dst_idx.end());
  93. // Store the indices and distances
  94. for (int i = 0; i < k; ++i) {
  95. idx(j, i) = dst_idx[i].second;
  96. dst(j, i) = dst_idx[i].first;
  97. }
  98. }
  99. }
  100. }
  101. #endif /* V0_HPP_ */