[SciPy-dev] Trouble with optimize.fmin_ncg

[Nils Wagner]

Stefan van der Walt wrote:

Stefan,

Thank you very much for running the test.
Is there an easy way to check the differences between r3116 and r3189
wrt. scipy.optimize ?

Nils

P.S I have send two additional tests to Alan Isaac off-list. They
exhibit similar "failures" of ncg.


> The optimisation converges under scipy r3116:
>
> $ python test_tnc.py 
> try fmin_bfgs
> Optimization terminated successfully.
>          Current function value: 0.000540
>          Iterations: 263
>          Function evaluations: 266
>          Gradient evaluations: 266
> Rayleigh quotient BFGS 0.000540272922563
> fmin_bfgs OK
> try fmin_ncg
> Optimization terminated successfully.
>          Current function value: 0.000538
>          Iterations: 13
>          Function evaluations: 33
>          Gradient evaluations: 13
>          Hessian evaluations: 13
> Rayleigh quotient NCG 0.000537952836927
>
> Regards
> Stéfan
>
> On Tue, Jul 24, 2007 at 11:12:23PM -0400, Alan G Isaac wrote:
>   
>> On Tue, 24 Jul 2007, dmitrey apparently wrote:
>>     
>>> can anyone who didn't update scipy from svn during last 
>>> 10-11 days run Nils example? 
>>>       
>> The core of Nil's example is below.
>> Get the data from here:
>> http://math.nist.gov/MatrixMarket/data/Harwell-Boeing/lanpro/nos4.html
>> Cheers,
>> Alan Isaac
>>     
> _______________________________________________
> Scipy-dev mailing list
> Scipy-dev at scipy.org
> http://projects.scipy.org/mailman/listinfo/scipy-dev
>