[SciPy-User] optimize.fmin_l_bfgs_b wrong number of arguments

josef.pktd@gmai... josef.pktd@gmai...
Sun Jul 22 11:30:08 CDT 2012


On Sun, Jul 22, 2012 at 12:22 PM, ilan barak <ibarak2000@yahoo.com> wrote:
> Hello,
> I apologize for the long description, this is the best I can do...
>
> I have two functions defined as:
>
> def hypothesis(params):
>     # build a series of length params[0], with starting gap
>     params[1] , repitition params[2] with sclae params[3]
>     # using tt as the basic shape
>     result=np.zeros(params[0])
>     for index in
>     np.arange(params[1],params[0]-tt.shape[0],params[2]):
>         result[int(round(index)):int(round(index))+tt.shape[0]]=tt
>     return params[3]*result
>
> def Cost(params,sig): #  starting gap params[0] , repitition
>     params[1] with sclae params[2]
>     # sig is signa;s to compare to
>     # calculate error cost
>     result=np.linalg.norm(sig-hypothesis([sig.shape[0]]+params))
>     return result
>
> The Cost function requires a 3 parameter list and a signal that is a
>     800 long ndarray
> Running the Cost with:
> params=([start,gap,0.04])
> Cost(params,mysig5) , where mysig 5 is length 800 ndarray works
>     fine.
> However:
> p0 = np.array([10.,20.,0.01]) # Initial guess for the parameters
>     start,gap,scale
> mybounds = [(0,20), (10,25), (0.001,0.1)]
> x, f, d= optimize.fmin_l_bfgs_b(Cost, p0[:],fprime=None,
>     args=mysig5, bounds=mybounds, approx_grad=True)

args=(mysig5,)

args should be a tuple

I think

Josef
>
> complains:
>
> Traceback
>     <module>
>     C:\Users\ilan\Documents\python\opt_detection_danny1.py    130
>     fmin_l_bfgs_b
>     C:\Python27\lib\site-packages\scipy\optimize\lbfgsb.py    199
>     func_and_grad
>     C:\Python27\lib\site-packages\scipy\optimize\lbfgsb.py    145
> TypeError: Cost() takes exactly 2 arguments (801 given)
>
>
> Where am I wrong
>
> thanks
>
> Ilan
> _______________________________________________
> SciPy-User mailing list
> SciPy-User@scipy.org
> http://mail.scipy.org/mailman/listinfo/scipy-user


More information about the SciPy-User mailing list