[SciPy-user] argmin and optimization

Alan G Isaac aisaac@american....
Thu Jul 26 07:18:53 CDT 2007


On Thu, 26 Jul 2007, Nils Wagner apparently wrote:
> I would like to implement an algorithm described by Golub and Liao. 
> Get the source from here: http://citeseer.ist.psu.edu/638954.html 

> The central question is the following. How do I implement the operator 
> P_\Omega(\cdot)  in scipy ? 

> The definition is as follows (\LaTex notation)

> P_\Omega(y) = \textrm{argmin}\limits_{x \in \Omega} \|x-y\|_2 \forall \, 
> y \in \mathds{R}^n 

> \Omega = \{ x \in \mathds{R}^n | <x,x> \le 1 \}. 


It looks like you are saying:
get as close to y as possible without leaving the unit 
sphere.  Is that right?  This is just y if y is in the
unit sphere and it is y normalized to unit length if
y is outside the unit sphere.

Please post the code when you finish the algorithm.

Cheers,
Alan Isaac




More information about the SciPy-user mailing list