Julian Panetta
Julian Panetta
You should be able to call GeneralizedEigenSolver::eigenvectors(); this method doesn't appear in the documentation, but it exists. It seems it's [still an open problem](http://www.ma.man.ac.uk/~ftisseur/talks/iwasep.pdf) to develop an efficient algorithm that...
No idea... the following compiles for me with Eigen 3.3.4: ``` #include #include #include int main(int argc, const char *argv[]) { Eigen::MatrixXd A, B; A.setIdentity(3, 3); B.setIdentity(3, 3); Eigen::GeneralizedEigenSolver ges;...
Especially for such a small problem, I don't think there's anything wrong with transforming it into an ordinary, asymmetric eigenvalue problem as you wrote. Note that C^-1 just exchanges the...
Are you trying to add the edges from the solve() function? If so, they're probably being overridden by callback_pre_draw. You should add your visualization code to callback_pre_draw instead.
Weird, I've not seen this happen--does it occur on a freshly checked out copy?
Sorry for the delayed reply. Indeed L is the *symmetric* cotan Laplacian (holding the cotan weights only, with no division by the averaging region area). M is the mass matrix,...
This doesn't seem to be what you want. You should look into [vec.replicate](https://eigen.tuxfamily.org/dox/classEigen_1_1DenseBase.html#a4f89ab26ccf66c3a9d0190e10243cb9e) and also [vec.asDiagonal()](https://eigen.tuxfamily.org/dox/classEigen_1_1MatrixBase.html#ab757d5801a0e038f8555635f06456352).
The slides were written assuming the variables were ordered so that all free variables appear first followed by all constrained variables. Then A can be written as the block matrix...
Yes, for the case of matrix G^T G, row i corresponds to vertex i and column j to vertex j. The ith row of G^T G u is the value...
I think this should be straightforward if you understand where G^T G comes from. Pretend for the moment that the energy doesn't have area weights: E_1(s) := sum_t ||g_t -...