[SciPy-user] leastsq information message

zunzun@zunzu... zunzun@zunzu...
Thu Apr 19 11:58:01 CDT 2007


On Thu, Apr 19, 2007 at 09:05:40AM -0400, Tommy Grav wrote:
> 
> Both actual and predicted relative reductions in the sum of squares
>    are at most 0.000000

I poked around and found in the scipy source code minpack.py:

(lmdif or lmder is first called and this test is made)
if info in [5,6,7,8]:
    print "Warning: " + errors[info][0]


and according to the minpack lmdif info available at:

http://www.math.utah.edu/software/minpack/minpack/lmdif.html

one of the following (5,6,7,8) causes your error message to print

         INFO = 5  Number of calls to FCN has reached or exceeded
                   MAXFEV.

         INFO = 6  FTOL is too small.  No further reduction in the sum
                   of squares is possible.

         INFO = 7  XTOL is too small.  No further improvement in the
                   approximate solution X is possible.

         INFO = 8  GTOL is too small.  FVEC is orthogonal to the
                   columns of the Jacobian to machine precision.

So it looks to be caused by one of these conditions.

     James Phillips
     http://zunzun.com


More information about the SciPy-user mailing list