variables changing after a function call in python

Question

Ive been working on some code recently for a project in python and im confused by the output im getting from this code

def sigmoid(input_matrix):
    rows = input_matrix.shape[0]
    columns = input_matrix.shape[1]
    for i in range(0,rows):
        for j in range(0,columns):
            input_matrix[i,j] = (1 / (1 + math.exp(-(input_matrix[i,j]))))
    return input_matrix

def feed_forward(W_L_1 , A_L_0 , B_L_1):
    weighted_sum = np.add(np.dot(W_L_1,A_L_0), B_L_1)
    activation = sigmoid(weighted_sum)
    return [weighted_sum,activation]

a = np.zeros((1,1))
b = feed_forward(a,a,a)
print(b[0])
print(b[1])

when I print both b[0] and b[1] give values of .5 even though b[0] should equal 0. Also in addition to this when I place the '''weighted_sum = np.add(np.dot(W_L_1,A_L_0), B_L_1)''' again after the 'actvation' line it provides the correct result. Its as if the 'activation' line has changed the value of the weighted sum value. Was wondering if anyone could spread some light on this I can work around this but am interested as to why this is happening. Thanks!!

your sigmoid function changes the array you pass to it. input_matrix[i,j] = ... why did you expect it not to change — juanpa.arrivillaga
– juanpa.arrivillaga, Commented Jan 22, 2021 at 16:18

Óscar López · Accepted Answer · 2021-01-22 16:16:49Z

1

Inside sigmoid, you're changing the value of the matrix passed as parameter, in this line:

input_matrix[i,j] = ...

If you want to prevent this from happening, create a copy of the matrix before calling sigmoid, and call it like this: sigmoid(copy_of_weighted_sum).

answered Jan 22, 2021 at 16:16

Óscar López

237k38 gold badges321 silver badges391 bronze badges

Sign up to request clarification or add additional context in comments.

6 Comments

Joth Over a year ago

Thanks for the idea I've gave this a try but still got the same result!

Óscar López Over a year ago

My recommendation has to work. How did you copy the matrix?

Joth Over a year ago

def sigmoid(input_matrix):     test = input_matrix     rows = test.shape[0]     columns = test.shape[1]     for i in range(0,rows):         for j in range(0,columns):             test[i,j] = (1 / (1 + math.exp(-(input_matrix[i,j]))))     return test

sorry about the formatting

Joth Over a year ago

I was under the impression too and I could be very wrong about this that the scope of the variable was only within that function and wasnt passed down to an inner function ie:def inner(a): a = 3 def outer(): a = 4 inner(a) print(a) outer()gives 4 as the result

Óscar López Over a year ago

Here's the problem, this line is not copying the matrix, at all! test = input_matrix. That's just creating another reference pointing to the exact, same object. You've been passing around and modifying a single matrix, and of course any change you do on it will be reflected on all the variables referencing it.

|

Collectives™ on Stack Overflow

variables changing after a function call in python

1 Answer 1

6 Comments

Your Answer

Linked

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

6 Comments

Your Answer

Sign up or log in

Post as a guest

Linked

Related