How should I initialize the contents of a large matrix in Eigen?

Question

I am trying to initialize a matrix (using the Eigen library) to have a nonzero value when I create it. Is there a nice way to do this without a for loop?

For example, if I wanted to initialize the whole matrix to 1.0, I would like to do something like:

Eigen::MatrixXd mat(i,j) = 1.0;

or

Eigen::MatrixXd mat(i,j);
mat += 1.0;

(I am used to this type of thing in MATLAB, and it would make Eigen even nicer to use than it already is. I suspect there is a built-in method somewhere that does this, that I have not found.)

A sub-question to this question would be how to set a block of matrix elements to a set value, something ilke:

mat.block(i,j,k,l) = 1.0;

Am

I found an answer, but it would still be nice to have a syntax like I proposed... :) – Am

Close to what you want: multiply the scalar by Eigen::MatrixXd::Ones(rows,cols), like: Eigen::MatrixXd mat(3,3) = 1.5 * Eigen::MatrixXd::Ones(3,3) It's not quite like MATLAB, but close – Thenna

The syntax you tried works with Eigen::Array but not with linear algebra matrices because in this case a scalar value should rather be assimilated as the identity matrix times this scalar value. – Bradbradan

Answer 1

As so often happens I found the answer in the docs within thirty seconds of posting the question. I was looking for the Constant function:

Eigen::MatrixXd mat = Eigen::MatrixXd::Constant(i, j, 1.0);

mat.block(i,j,k,l) = Eigen::MatrixXd::Constant(k, l 1.0);

Am

There are also other "init" functions, like Eigen::Matrix::Zero() or Eigen::Matrix::Ones() eigen.tuxfamily.org/dox/… – Thenna

As well as the setConstant(val), setOnes(), setZero(), fill(val) variants which are more compact to write with less redundancy in your cases. – Bradbradan

Be cautious of the parameters for block and Eigen::MatrixXd::Constant. ( i, j ) is the starting position of the block, and ( k, l ) is the size of the block. So, I think in the second example the ( i, j ) pair on the right hand side should be ( k,l ). – Bug

Are the variables i and j indices or the number of rows and columns? – Lester

@Lester It's the number of rows and columns. I just corrected my post based on jmcarter9t's comment. – Am

Answer 2

Eigen::MatrixXd::Ones(), Eigen::MatrixXd::Zero() and Eigen::MatrixXd::Random() can all give you what you want, creating the Matrix in a dynamic way.

Ldopa