Hello,

I have the following snippet where I'm computing the partial derivatives for solving the regularized ridge regression so need to avoid regularizing the first (bias) coefficient:

... + lambda*[0; thetas(2, end)]

Basically I create a new column vector by appending zero (first row) with the relevant theta elements, how can I achieve the same using Eigen in the most succinct way?

TIA,
Best regards,
Giovanni

... + lambda * (VectorXd(end-1) << 0, thetas.segment(2,end-2)).finish() ...

or create a tmp:

Code: Select all

VectorXd tmp(end-1);
tmp << 0, thetas.segment(2,end-2);

.. + lambda * tmp ...


You might also consider thetas.tail(end-2) if you want the last end-2 elements.

Very useful, thank you!