def __init__(self,learningrate, func, dfunc, Xinit, args, maxiter,outfilename):

GradientDescent.__init__(self,learningrate, func, dfunc, Xinit, args, maxiter,outfilename)

def minimize(self):

X = self.Xinit

tau = 1.

kappa = 0.5

for ii in range(self.maxiter):

indices=np.unique(self.args[1]).astype(np.uint32)

chosen_point = np.random.randint(indices.shape[0])

lrate_ii = self.lrate*pow(ii+tau,-kappa)

dX = self.dfunc(X,chosen_point,*self.args)

X -= lrate_ii * dX

if self.outfilename == 'none':

xxx=0

elif self.outfilename == sys.stdout:

f = self.outfilename

else:

f = open(self.outfilename,'w')

loss = self.func(X,chosen_point,*self.args)

if self.outfilename != 'none':

f.write(' Loss %6i; Value %.3f\r' % (ii, loss))

if self.outfilename != 'none' and self.outfilename != sys.stdout:

f.close()