[Scipy-tickets] [SciPy] #728: maxiter broken for Scipy.sparse.linalg gmres, in addition to no ability to track convergence

SciPy scipy-tickets@scipy....
Tue Aug 26 11:15:14 CDT 2008


#728: maxiter broken for Scipy.sparse.linalg  gmres, in addition to no ability to
track convergence
----------------------+-----------------------------------------------------
 Reporter:  jschrod3  |        Owner:  somebody
     Type:  defect    |       Status:  new     
 Priority:  normal    |    Milestone:          
Component:  Other     |      Version:          
 Severity:  normal    |   Resolution:          
 Keywords:            |  
----------------------+-----------------------------------------------------
Comment (by jschrod3):

 Ticket 360 does address the problem of tracking convergence.  It does not
 address the problem of the maxiter variable not functioning in gmres.  The
 maxiter problem is, however, very easy to fix.  Incrementing a variable
 every time the residual is updated isn't hard.

 The point of my ticket, is to request that the official scipy be updated
 so that gmres does use maxiter and does allow for tracking of convergence.
 As it stands now, when a user (like me), types "gmres?" in ipython, he's
 told that the maxiter and callback parameters work.  That means that users
 end up spending quite a bit of time trying to figure out what they're
 doing wrong.  A typical user is unlikely to think that these features
 aren't implemented, as these are basic functionalities of an iterative
 method, especially a workhorse like gmres.  At the very least, the
 comments documenting gmres should be changed so that users know that
 callback and maxiter don't work.

 Its true that with ticket 360 and little thought, the shortcomings of the
 current gmres can be overcome, but that defeats the purpose of having a
 package like scipy that's supposed to obviate reinvention of the wheel.

 Lastly in the future, instead of creating a new ticket for something like
 this, I'll just work with the existing ticket (assuming one exists) that
 pertains to most of my points.  So, I apologize for creating a new ticket
 instead of working with 360.  I belatedly realize that its bad form.

-- 
Ticket URL: <http://scipy.org/scipy/scipy/ticket/728#comment:3>
SciPy <http://www.scipy.org/>
SciPy is open-source software for mathematics, science, and engineering.


More information about the Scipy-tickets mailing list