matrix multiplication with Rcpp - different values when the output is assigned

Question

I wrote the following function to calculate a matrix (information matrix of Weibull model)

#include <RcppEigen.h>
#include <math.h>
#include <vector>
using namespace std;

using Eigen::MatrixXd;                  

// [[Rcpp::depends(RcppEigen)]]
// [[Rcpp::export]]

MatrixXd Weibull_FIM(const vector<double> x, const vector<double> w, const vector<double> param)
{
  if(x.size() != w.size()){
    Rcpp::Rcout<<"The length of x and w is not equal."<<std::endl;
    exit(1);
  }
  double  a, b, lambda, h, constant;
  a = param[0];
  b = param[1];
  lambda = param[2];
  h = param[3];

  a = a + 0; //just to not get a warning 

  Eigen::MatrixXd Fisher_mat(4, 4);  
  size_t i;

  for(i=0; i < x.size(); i++)
  { 
    constant = exp(-lambda * pow(x[i], h));
    Eigen::MatrixXd f(4, 1);
    f(0, 0) = 1;
    f(1, 0) = -constant;
    f(2, 0) = b*pow(x[i], h)*constant;
    f(3, 0) = b*pow(x[i], h)*constant * lambda * log(x[i]);

    Fisher_mat = w[i] * f * f.transpose() + Fisher_mat;
  }

  return Fisher_mat;
}

Now I want to calculate the matrix for some values in R:

 Weibull_FIM(x=c(1, 1.239, 1.749371, 5), w = rep(.25, 4), param=c(1, 1, 1, 2))

#            [,1]         [,2]         [,3]         [,4]
#[1,]  1.00000000 -0.157545687  0.210509365  0.037774174
#[2,] -0.15754569  0.045985587 -0.053326010 -0.004757122
#[3,]  0.21050936 -0.053326010  0.066320481  0.008736574
#[4,]  0.03777417 -0.004757122  0.008736574  0.002864707

This matrix is alright. Now, if I first assign the output and then print it, I will have a different matrix (some of the elements will be changed to some large values!).

res <- Weibull_FIM(x=c(1, 1.239, 1.749371, 5), w = rep(.25, 4), param=c(1, 1, 1, 2))
print(res)
#           [,1]           [,2]         [,3]           [,4]
#[1,]  1.00000000  4.727161e+180  0.210509365  2.267126e+161
#[2,] -0.15754569  5.104678e+199 -0.053326010  -4.757122e-03
#[3,]  0.21050936   2.079498e+64  0.066320481   8.736574e-03
#[4,]  0.03777417  -4.757122e-03  0.008736574   2.864707e-03

As can be seen, elements (1, 2), (1, 4), (2, 2) and (3, 2) get another large values. I would appreciate any help.

Update1 in response to @Ronald:

• version.string R version 3.1.2 (2014-10-31)
• Windows 7 x64
• platform x86_64-w64-mingw32
• Rcpp_0.11.3, RcppEigen_0.3.2.3.0, tools_3.1.2
• Rtools version 3.2.0.1948

Answer 1

You forgot to initialize Fisher_mat with zeroes before summation, so it sometimes may contain random junk. Adding e.g. Fisher_mat.setZero(); before a for loop will give you a consistent output.

Check this reference: RcppEigen Introduction.