animatrix
patreon
VEX Singular Value Decomposition (SVD) in PCA SOP 2.0
🕑 Added 2020-02-17 13:00:01 +0000 UTCPCA SOP now uses SVD (Singular Value Decomposition) to compute the PCA, instead of EVD (Eigen Value Decomposition).
SVD is more precise, robust and numerically reliable than EVD.
The new version is completely redesigned, where the computed eigen vectors and the orient attribute will be on the points, and P will represent centroid (mean), unless disabled in the point cloud method.
Depending on the method used, it will either create a single point for the entire point cloud, for each cluster or keep the original points as is when doing a point cloud lookup.
Note that the new operator requires a very recent daily build of H18.
https://www.dropbox.com/s/izpm6lkjvb96wpq/PCA_SOP_Animatrix.otl?dl=1
