[Numpy-discussion] minimize Algorithmen Problem with boundarys

[Vertigo]

Hello everybody,
i have a Problem, with three minimize Algorithmen in scipy. I need
Algorithmen to support a minimize Algorithmen with boundarys. My favorits
are
the 'l-bfgs-b', 'slsqp' and 'tnc' but all of them give a failure meanwhile
solve my problem.
  
By call the methode SLSQP it give me failure back out of methode in file
slsqp.py in the line 397 of code: "failed in converting 8th argument 'g' of
_slsqp.slsqp to C/Fortran".

my Jacobi Matrix ist a differentiate from the jccurve(in the Code below)
formula and i code this
in Lambda formalism. The Variable "x" are the optimize Parameter in this
case
also x[0]=a,x[1]=b,x[2]=n,x[3]=c
In my case sigma and epsilon is the flow curve of metal and i optimize the
Johnson-Cook curve with this Code. That mean sigma ist a array with 128
elements also epsilon is a array with 128 elements. The Paramete a,b,n,c are
number in this case.

The Solver l-bfgs-b give me a failure back out of the methode in the file
lbfgsb.py in the line 264 in the function g=fac(x,*args). That function is a
referenz to my Jacobi-matrix "jcjac" and it give me a TypeError
"'builtin_function_or_method' object has no attribute '__getitem__'" I don´t
what that mean. This failure also in the Solver from TNC.

I hope i could explain my Problem and everbody can help me.

Here begin my extract from code:

jcjac = lambda x,sig,eps,peps : array[(1+x[3]*log(peps/1)),
((eps**x[2])+(eps**x[2])*log(peps/1)+ x[0]*log(peps/1)), (
x[0]*log(peps/1)),
(x[1]*(eps**x[2])*log(eps)+(eps**x[2])*log(eps)*log(peps/1))]

def __jccurve__(self,a,b,n,c,epsilon,pointepsilon):
		sig_jc=(a+b*epsilon**n)*(1+c*numpy.log(pointepsilon/1))
		return sig_jc
def __residuals__(self,const,sig,epsilon,pointepsilon):
		return
sig-self.__jccurve__(const[0],const[1],const[2],const[3],epsilon,pointepsilon)

p_guess=a1,b1,n1,c
max_a=a1+(a1*0.9)
min_a=a1-(a1*0.9)
min_a=0
max_b=b1+(b1*0.9)
min_b=b1-(b1*0.9)
min_b=0
max_n=n1+(n1*10)
min_n=n1-(n1*10)
min_n=0
max_c=c+(c*100)
min_c=c-(c*100)
min_c=0

optimize.minimize(self.__residuals__,p_guess,args=(sigma,epsilon,pointepsilon),method='slsqp',jac=jcjac,
hess=None,hessp=None,bounds=((min_a,max_a),(min_b,max_b),(min_n,max_n),(min_c,max_c)))

best greets
Lars




--
View this message in context: http://numpy-discussion.10968.n7.nabble.com/minimize-Algorithmen-Problem-with-boundarys-tp37709.html
Sent from the Numpy-discussion mailing list archive at Nabble.com.