[Scipy-tickets] [SciPy] #1605: Bug in scipy.interpolate.griddata (numerical problem)

SciPy Trac scipy-tickets@scipy....
Wed Feb 22 14:31:10 CST 2012


#1605: Bug in scipy.interpolate.griddata (numerical problem)
----------------------------------------+-----------------------------------
 Reporter:  schicketanz                 |       Owner:  somebody   
     Type:  defect                      |      Status:  new        
 Priority:  high                        |   Milestone:  Unscheduled
Component:  scipy.interpolate           |     Version:  0.10.0     
 Keywords:  griddata numerical problem  |  
----------------------------------------+-----------------------------------
Changes (by pv):

  * priority:  normal => high


Comment:

 This indeed seems to be a precision error -- for this input data, the
 triangulation qhull produces contains nearly (but not completely)
 degenerate triangles, for which the computed barycentric transforms are
 nearly singular. This then confuses
 `scipy.spatial.qhull._find_simplex_directed` to thinking that certain
 points lie outside the convex hull, although in reality they don't.

 This doesn't seem to occur on all machines, probably depends on compiler
 etc.

 The solution is to check the validity of the transform, and discard the
 simplices which appear to be degenerate. Here's a fix:

 - https://github.com/pv/scipy-work/commits/bug/1605-griddata-baryfix

-- 
Ticket URL: <http://projects.scipy.org/scipy/ticket/1605#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