[Scipy-tickets] [SciPy] #1727: LinearNDInterpolator hangs on 4D data set

SciPy Trac scipy-tickets@scipy....
Sun Sep 30 10:42:04 CDT 2012


#1727: LinearNDInterpolator hangs on 4D data set
--------------------------------+-------------------------------------------
  Reporter:  miked983           |       Owner:  pv                               
      Type:  defect             |      Status:  closed                           
  Priority:  high               |   Milestone:  Unscheduled                      
 Component:  scipy.interpolate  |     Version:  0.10.0                           
Resolution:  invalid            |    Keywords:  LinearNDInterpolator, interpolate
--------------------------------+-------------------------------------------
Changes (by pv):

  * status:  new => closed
  * resolution:  => invalid


Comment:

 The data set is too large for Qhull to triangulate in 4-D quickly. The
 computation will eventually finish, but this can take several hours (the
 scaling is probably somewhere between N^2 and N^3).

 That you can use `interpn` means your data is not scattered, but on a
 rectangular grid. `interpn` is based on a tensor-product interpolation,
 whereas `LinearNDInterpolator` (similarly to `griddata`), is
 triangulation-based scattered data interpolation. Using these for
 rectangular grid data is very inefficient, and you hit data set size
 limits way earlier.

 Scipy does not currently ship with an equivalent of `interpn`. However, it
 is very easy to construct from existing pieces:
 http://mail.scipy.org/pipermail/scipy-user/2012-June/032314.html

 Yes, improving rectangular grid data interpolation routines and adding the
 tensor product interpolation in Scipy would be nice, and it's on the TODO
 list and moreover fairly easy to implement.

-- 
Ticket URL: <http://projects.scipy.org/scipy/ticket/1727#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