Why not the second line in your code, usually the first one is not used in books or papers.
If kf. S is an orthogonal matrix, the results of the upper and lower rows are consistent,otherwise they are different.
The update of the a posteriori estimate covariance Pp was taken from the book "Simo Särkkä (2013). Bayesian Filtering and Smoothing. Cambridge University Press", Eq. 3.10.
3.kf_update.m kf.Pp = kf.Pi - kf.K kf.S kf.K'; %kf.Pp = kf.Pi - kf.K kf.H kf.Pi;
Why not the second line in your code, usually the first one is not used in books or papers. If kf. S is an orthogonal matrix, the results of the upper and lower rows are consistent,otherwise they are different.