[SciPy-dev] problem with linalg.cholesky?

[Jeff Whitaker]

Andrew Jaffe wrote:

>hi all,
>
>(apologies that a similar question has appeared elsewhere...)
>
>In the newest incarnation of scipy_core, I am having trouble with the
>cholesky(a) routine. Here is some minimal code reproducing the bug  
>(on OS X)
>
>------------------------------------------------------------
>from scipy import identity, __core_version__, Float64
>import scipy.linalg as la
>print 'Scipy version: ', __core_version__
>i = identity(4, Float64)
>print 'identity matrix:'
>print i
>
>print 'about to get cholesky decomposition'
>c = la.cholesky(i)
>print c
>------------------------------------------------------------
>
>which gives
>
>------------------------------------------------------------
>Scipy version:  0.6.1
>identity matrix:
>[[ 1.  0.  0.  0.]
>  [ 0.  1.  0.  0.]
>  [ 0.  0.  1.  0.]
>  [ 0.  0.  0.  1.]]
>about to get cholesky decomposition
>Traceback (most recent call last):
>   File "/Users/jaffe/Desktop/bad_cholesky.py", line 13, in ?
>     c = la.cholesky(i)
>   File "/Library/Frameworks/Python.framework/Versions/2.4/lib/ 
>python2.4/site-packages/scipy/linalg/basic_lite.py", line 117, in  
>cholesky_decomposition
>     results = lapack_routine('L', n, a, m, 0)
>lapack_lite.LapackError: Parameter a is not contiguous in  
>lapack_lite.dpotrf
>------------------------------------------------------------
>
>(The cholesky decomposition in this case should just be the matrix  
>itself; the same error occurs with a complex matrix.)
>
>
>Any ideas? Could this have anything to do with _CastCopyAndtranspose  
>in Basic_lite.py? (Since there are few other candidates for anything  
>that actually changes the matrix.)
>
>Thanks in advance,
>
>A
>
>  
>

Another data point - on OS X 10.3.9 with python 2.4.2, scipy_core 0.6.1 
and scipy 0.4.3 I get the expected result (no error).

-Jeff

-- 
Jeffrey S. Whitaker         Phone  : (303)497-6313
Meteorologist               FAX    : (303)497-6449
NOAA/OAR/PSD  R/PSD1        Email  : Jeffrey.S.Whitaker at noaa.gov
325 Broadway                Office : Skaggs Research Cntr 1D-124
Boulder, CO, USA 80303-3328 Web    : http://tinyurl.com/5telg