Question

给定一个数据矩阵X，我想计算任意两行X之间的成对距离矩阵。我有以下代码，它来自于稍微调整代码here。

#include <Rcpp.h>
#include <cmath>
#include <algorithm>

using namespace Rcpp;
// generic function for l1_distance
template <typename InputIterator1, typename InputIterator2>
inline double l1_distance(InputIterator1 begin1, InputIterator1 end1, 
                            InputIterator2 begin2) {
   double rval = 0;
   InputIterator1 it1 = begin1;
   InputIterator2 it2 = begin2;
   while (it1 != end1) {
      double d1 = *it1++;
      double d2 = *it2++;
      rval += abs(d1 - d2);
   }
   return rval;  
}

// [[Rcpp::export]]
NumericMatrix rcpp_l1_distance(NumericMatrix mat) {

   // allocate the matrix we will return
   NumericMatrix rmat(mat.nrow(), mat.nrow());
   for (int i = 0; i < rmat.nrow(); i++) {
      for (int j = 0; j < i; j++) {
         NumericMatrix::Row row1 = mat.row(i);
         NumericMatrix::Row row2 = mat.row(j);
         double d = l1_distance(row1.begin(), row1.end(), row2.begin());
         rmat(i,j) = d;
         rmat(j,i) = d;
      }
   }
   return rmat;
}

问题是此代码返回所有整数值的矩阵。整数值似乎与我想要的距离值正相关，这使得它更加混乱。我还计算了成对的l2距离矩阵和成对归一化的l1距离（将两行之间的l1距离除以它们的l1范数之和）矩阵，并且它们都表现得如预期的那样。

有人能告诉我哪个部分犯了错误吗？

您可以执行以下操作以获得奇怪的结果

library(Rcpp)
sourceCpp("distance.cpp") #the file containing the cpp code above
X = matrix(rnorm(16), 4, 4)
rcpp_l1_distance(X)

提前致谢！

Answer 1

编译代码会给我这些警告：

> Rcpp::sourceCpp('Desktop/mat.cpp')
mat.cpp:16:15: warning: using integer absolute value function 'abs' when argument is of floating point type [-Wabsolute-value]
      rval += abs(d1 - d2);
              ^
mat.cpp:16:15: note: use function 'std::abs' instead
      rval += abs(d1 - d2);
              ^~~
              std::abs
mat.cpp:16:15: warning: using integer absolute value function 'abs' when argument is of floating point type [-Wabsolute-value]
      rval += abs(d1 - d2);
              ^
mat.cpp:30:18: note: in instantiation of function template specialization 'l1_distance<Rcpp::MatrixRow<14>::iterator, Rcpp::MatrixRow<14>::iterator>' requested here
      double d = l1_distance(row1.begin(), row1.end(), row2.begin());
                 ^
mat.cpp:16:15: note: use function 'std::abs' instead
      rval += abs(d1 - d2);
              ^~~
              std::abs
2 warnings generated.

...暗示abs适用于整数，请参阅this help page，您可以使用fabs或std::abs，或者您可以使用糖operator-，abs和sum：

#include <Rcpp.h>

using namespace Rcpp;

// [[Rcpp::export]]
NumericMatrix rcpp_l1_distance(NumericMatrix mat) {

  // allocate the matrix we will return
  NumericMatrix rmat(mat.nrow(), mat.nrow());
  for (int i = 0; i < rmat.nrow(); i++) {
    NumericMatrix::Row row1 = mat.row(i);

    for (int j = 0; j < i; j++) {
      rmat(j,i) = rmat(i,j) = sum( abs(row1 - mat.row(j) )) ;
    }
  }
  return rmat;
}

为什么我的Rcpp代码会产生意外的所有整数值输出？

1 个答案: