0

I am creating a simple neural network that uses an image as it's input. It's all done except that I have to minimize my cost function, but when I run the scipy.optimize.minimize, it just sits there. I'd assume that it shouldn't take long as the amount of data I'm passing through isn't very large. My question is why won't it terminate successfully and why it doesn't throw up a an error instead. It is also using almost all of my cpu usage after running the program, but hours later it still won't terminate.

import numpy

input_layer_size = 625

hidden_layer_size = 40

num_labels = 1

Theta1 = randInitializeWeights(input_layer_size, hidden_layer_size)

Theta2 = randInitializeWeights(hidden_layer_size, num_labels)

nn_params = list(Theta1.flat) + list(Theta2.flat)

numpy.asarray(nn_params)



def CostFunction(nn_params) # as well as other values(args)

    #Cost Function that unrolls nn_params and return an int (J)

from scipy import optimize

res = scipy.optimize.minimize(costFunction,x0 = nn_params,args = (input_layer_size, hidden_layer_size,num_labels,X,y,lambd),options = {'maxiter':50,'disp':True})

print(res)
Sandesh Ramesh
  • 1
  • 2
  • What kind of help do you expect? Have you tried this option with a simpler example? – hpaulj Aug 12 '19 at 17:12
  • Sorry I just realized the display option is text that appears after the minimization has terminated. (I will edit my question) but I still don't understand why my code won't terminate successfully, and yes I have tried it with the rosen cost function and it works fine. Should I copy my whole cost function into the code portion of question, or rather does it have to do with the way I called the minimization function? – Sandesh Ramesh Aug 12 '19 at 20:29

0 Answers0