0

Is there any built-in way in Eigen to re-orthogonalize matrix? When you multiply lots of rotations, matrix will eventually need to be re-orthogonalize. There are standard techniques such as using SVD and one can certainly spend a day writing and optimizing it by hand. However I'm hoping Eigen has few things built-in somewhere. There was a related question asked before but without clear answer in context of Eigen.

Community
  • 1
  • 1
Shital Shah
  • 63,284
  • 17
  • 238
  • 185
  • There is no clear answer in the previous related question because the right answer depends on your application and needs, ranging from [Ali's answer](http://stackoverflow.com/a/23082112/1641621) if the matrix is already nearly orthogonal, to a full polar-decomposition based on SVD (see [class Transform](http://eigen.tuxfamily.org/dox-devel/classEigen_1_1Transform.html#aeea08b29b4e9d4e02c8f3acdd4eabd1f) ) if you want the closest rotation in terms of L2 norm. – ggael Dec 30 '16 at 09:39
  • I'm not sure how exactly your question differs from the one you linked to. There is still no built-in re-orthogonalize method (but of course QR or SVD decompositions are built-in). If you want a specific method to be added, you can post a feature request in Eigen's bugtracker: http://eigen.tuxfamily.org/bz/ – chtz Dec 30 '16 at 18:39
  • Possible duplicate of [Eigen - Re-orthogonalization of Rotation Matrix](http://stackoverflow.com/questions/23080791/eigen-re-orthogonalization-of-rotation-matrix) – chtz Dec 30 '16 at 18:40

0 Answers0