I am currently porting some matlab code to C++ and trying to use the Eigen library. I have a lot of matrix-vector multiplications that need to work on sub-matrices. For example, in matlab one can have do referencing as an example as follows:

Code: Select all

U(:, 1:n)

Here, I have a matrix U where I am selecting all the rows and 'n' columns where 'n' is only known during run-time. Is it possible to generate such a view into the data using Eigen efficiently?

The line that I need to port is something like this in matlab:

Code: Select all

A_hat = U(:, 1:n) * diag(diagS(1:n) - 1/mu) * V(:, 1:n)'


Here U and V are matrices and diagS is a vector which gets turned into a diagonal matrix using the diag() function in matlab.

I realised there was a non-templates version of the block method. I have not tested it yet but I could do something like:

Code: Select all

X = U.block(0, 0, U.rows(), n) *
      singular_values.head(n).asDiagonal() *
      V.block(0, 0, V.rows(), n).transpose();


There are also convenient shortcuts, like U.leftCols(n). See them all there: http://eigen.tuxfamily.org/dox-devel/gr ... tions.html