Quote:
Originally Posted by Bongle
I believe that's the technique the matrix solving library I use (JAMA) uses. I don't know too many details about it.
|
For coefficient matrices associated with processing data from single events one at a time (even large events), LU is probably fine. But if you ever want to do computations for, say, an entire year's worth of data in one matrix (simultaneous solution using data from all events in the year), you might want to consider using Cholesky instead. It's twice as fast. The JAMA library you are using supports Cholesky.