Ticket #1161 (closed task: fixed)
|Reported by:||lanz||Owned by:||anze|
Anze (and Matija?), here is a patch that speeds up PCA 2-3x on large data sets with m>n. The implementation is also more straightforward and readable than the current magic transformations.
Since PCA is becoming so feature rich with degrees of freedom and generalized eig. options I am leaving it to you to work out any details this needs to fully conform. I just tested it on a normal case (script attached, run before and after patch).