Home | History | Annotate | Download | only in NonLinearOptimization
      1 namespace Eigen {
      2 
      3 namespace internal {
      4 
      5 template <typename Scalar>
      6 void covar(
      7         Matrix< Scalar, Dynamic, Dynamic > &r,
      8         const VectorXi &ipvt,
      9         Scalar tol = sqrt(NumTraits<Scalar>::epsilon()) )
     10 {
     11     typedef DenseIndex Index;
     12 
     13     /* Local variables */
     14     Index i, j, k, l, ii, jj;
     15     bool sing;
     16     Scalar temp;
     17 
     18     /* Function Body */
     19     const Index n = r.cols();
     20     const Scalar tolr = tol * abs(r(0,0));
     21     Matrix< Scalar, Dynamic, 1 > wa(n);
     22     assert(ipvt.size()==n);
     23 
     24     /* form the inverse of r in the full upper triangle of r. */
     25     l = -1;
     26     for (k = 0; k < n; ++k)
     27         if (abs(r(k,k)) > tolr) {
     28             r(k,k) = 1. / r(k,k);
     29             for (j = 0; j <= k-1; ++j) {
     30                 temp = r(k,k) * r(j,k);
     31                 r(j,k) = 0.;
     32                 r.col(k).head(j+1) -= r.col(j).head(j+1) * temp;
     33             }
     34             l = k;
     35         }
     36 
     37     /* form the full upper triangle of the inverse of (r transpose)*r */
     38     /* in the full upper triangle of r. */
     39     for (k = 0; k <= l; ++k) {
     40         for (j = 0; j <= k-1; ++j)
     41             r.col(j).head(j+1) += r.col(k).head(j+1) * r(j,k);
     42         r.col(k).head(k+1) *= r(k,k);
     43     }
     44 
     45     /* form the full lower triangle of the covariance matrix */
     46     /* in the strict lower triangle of r and in wa. */
     47     for (j = 0; j < n; ++j) {
     48         jj = ipvt[j];
     49         sing = j > l;
     50         for (i = 0; i <= j; ++i) {
     51             if (sing)
     52                 r(i,j) = 0.;
     53             ii = ipvt[i];
     54             if (ii > jj)
     55                 r(ii,jj) = r(i,j);
     56             if (ii < jj)
     57                 r(jj,ii) = r(i,j);
     58         }
     59         wa[jj] = r(j,j);
     60     }
     61 
     62     /* symmetrize the covariance matrix in r. */
     63     r.topLeftCorner(n,n).template triangularView<StrictlyUpper>() = r.topLeftCorner(n,n).transpose();
     64     r.diagonal() = wa;
     65 }
     66 
     67 } // end namespace internal
     68 
     69 } // end namespace Eigen
     70