Le mercredi 23 août 2006 à 21:35 +0200, Vanuxem Grégory a écrit :
[...]
I have just read the following code, you can greatly improve it :-)
(In fact it's a good thing that this code can be improved since I will
publish it one day (I don't know under which terms though)).
Greg
> =head2 stdize
>
> =for ref
>
> Standardization (possibly weighted) of matrix over specified axis:
>
> a - mean
> STANDARDIZED = ------------
> stdev(n-1)
>
> Uses arithmetic mean and standard deviation estimation if asked.
> Can use arbitrary values (PDL vector) and compute inplace.
>
> =for usage
>
> PDL = stdize(PDL, SCALAR(axis), SCALAR|PDL(center), SCALAR|PDL(scale),
> SCALAR(weight))
> axis : threading's axis, generally observation, default = 1 unless a
> vector is given
> center : center data by variable NOCENTER = 0 | CENTER = 1 | PDL,
> DEFAULT = 1
> scale : scale data by variable NOSCALE = 0 | SCALE = 1 | PDL, DEFAULT =
> 0
> weight : PDL of weights (size(entry matrix)), DEFAULT = ones(entry
> matrix)
>
> =for example
>
> my $a = random(10,10);
> my $standardized = stdize($a,1,1,1);
>
> =cut
>
> *stdize = \&PDL::stdize;
>
> sub PDL::stdize{
> my($m, $obs, $center, $scale, $weight) = @_;
> $obs = 1 unless defined($obs) || $m->getndims < 2;
> $center = 1 unless defined $center;
> my ($mean, $rms, $mm);
>
> $m = $m->copy unless $m->is_inplace(0);
>
> $mm = $m->mv($obs,0);
> if ( !UNIVERSAL::isa($center,'PDL') || !
> UNIVERSAL::isa($scale,'PDL')) {
> ($mean, $rms) = $mm->statsover($weight);
> $center = $mean if (!UNIVERSAL::isa($center,'PDL') &&
> $center);
> $scale = $rms if (!UNIVERSAL::isa($scale,'PDL') &&
> $scale);
> }
> $mm = $mm->mv(0,$m->getndims-1);
> if (UNIVERSAL::isa($center,'PDL')){
> $mm -= $center;
> }
> if (UNIVERSAL::isa($scale,'PDL')){
> $mm /= $scale;
> }
> $m;
>
> }
>
___________________________________________________________________________
Découvrez un nouveau moyen de poser toutes vos questions quelque soit le sujet
!
Yahoo! Questions/Réponses pour partager vos connaissances, vos opinions et vos
expériences.
http://fr.answers.yahoo.com
_______________________________________________
Perldl mailing list
[email protected]
http://mailman.jach.hawaii.edu/mailman/listinfo/perldl
