[Numpy-discussion] Efficient orthogonalisation with scipy/numpy

Gael Varoquaux gael.varoquaux@normalesup....
Tue Jan 19 15:45:34 CST 2010


On Tue, Jan 19, 2010 at 03:43:25PM -0600, Robert Kern wrote:
> On Tue, Jan 19, 2010 at 15:29, Charles R Harris
> <charlesr.harris@gmail.com> wrote:

> > Column stack the bunch so that the last column is y, then do a qr
> > decomposition. The last column of q is the (normalized) orthogonal vector
> > and its amplitude is the last (bottom right) component of r.

> Is the order actually guaranteed? In a quick test, it seems to work.

> In any case, I suspect that needing to do both x and y will make doing
> the QR once and some two pairs of dot products a better proposition
> than two QR decompositons.

Yes, that's correct, but my initial question wasn't clear enough, and
Chuck was answering my initial question.

Thanks a lot Chuck,

Gaël


More information about the NumPy-Discussion mailing list