[Scipy-tickets] [SciPy] #1666: LinearNDInterpolator fails when dimensions differ by many orders of magnitude

SciPy Trac scipy-tickets@scipy....
Sun Jun 3 07:36:28 CDT 2012


#1666: LinearNDInterpolator fails when dimensions differ by many orders of
magnitude
-----------------------------------------+----------------------------------
 Reporter:  bloop369                     |       Owner:  somebody   
     Type:  enhancement                  |      Status:  new        
 Priority:  low                          |   Milestone:  Unscheduled
Component:  scipy.interpolate            |     Version:  0.10.0     
 Keywords:  LinearNDInterpolator, QHull  |  
-----------------------------------------+----------------------------------
Changes (by bloop369):

 * cc: steven.byrnes@… (added)
  * priority:  normal => low
  * type:  defect => enhancement


Comment:

 OK, I changed this to the "enhancement" category. My request is to have a
 non-default option, maybe called rescale->True, that would scale the
 points to fill the unit n-dimensional box before running QHull. People
 would want to use this feature when some of the input dimensions have
 incommensurable units, like a function of space and time and light-
 intensity and bodyweight etc.

 I don't think more fine-grained control is necessary, because QHull seems
 very forgiving of different point-spacings along different dimensions.
 Even twelve orders of magnitude was tolerable in the example I gave above.
 It is not infinitely forgiving, but it is still very forgiving.

 Again, as far as I can tell, my request seems to have an incredibly simple
 implementation: Just set the 'QbB' flag for QHull. See
 http://www.qhull.org/html/qh-optq.htm#QbB ...

-- 
Ticket URL: <http://projects.scipy.org/scipy/ticket/1666#comment:2>
SciPy <http://www.scipy.org>
SciPy is open-source software for mathematics, science, and engineering.


More information about the Scipy-tickets mailing list