[Numpy-discussion] Polynomial implementation in numpy.ma

Stéfan van der Walt stefan@sun.ac...
Sat Sep 17 14:04:04 CDT 2011


On Sat, Sep 17, 2011 at 9:40 AM, Charles R Harris
<charlesr.harris@gmail.com> wrote:
> The Vandermonde matrix needs to be used for the fitting so nothing should be
> changed there.

I remember now where I heard this:

http://people.maths.ox.ac.uk/trefethen/mythstalk.pdf

Specifically, the statement: "Some well-known algorithms take O(n^2)
operations (e.g. naïve Lagrange formula)
or are exponentially unstable (e.g. Vandermonde matrices based on
monomials x^k )."

I wasn't sure whether to interpret that as the Vandermonde-based
algorithm being flawed, or the polynomial fitting problem on an
equidistant grid being ill-conditioned in general.

Regards
Stéfan


More information about the NumPy-Discussion mailing list