Python function behaves differently when i call it in another function loop

Question

I'm working on a machine-learning code that calculate the cost function and gradient descent , I wrote every function separately as shown:

def costFunction(theta, X, y):
    
    m = y.size

    J = (1/m) * ( np.dot(-y,np.log(sigmoid(np.dot(X,theta))))  -  np.dot((1-y),(np.log(1-sigmoid(np.dot(X,theta))))) ) 
    
    return J

def gradiantDescent(alpha , theta , X , y , num_itr):

    m = y.shape[0]
    J_history = []
    theta = theta.copy()

    for _ in range(num_itr):
        
        tempZero = theta[0]

        theta -= (alpha/m) * (np.dot(X.T , (sigmoid(np.dot(X,theta))-y)))
        theta[0] = tempZero -  ( (alpha/m) * np.sum((sigmoid(np.dot(X,theta))-y)))

        J_history.append(costFunction(theta, X, y))

    return theta , J_history

and when i call the 'cost function' separately it's works as i expected:

intial_theta = np.zeros(X.shape[1])

J = costFunction(intial_theta, X, y):

print(J) # works as expected

but when i call it in gradiantDescent function all J_history will be 'nan' value:

theta , Jvec = gradiantDescent(0.05, intial_theta , X , y , 500)

print(Jvec) #all values are 'nan'

So how can i fix it.

when you call gradiantDescent do you call costFunction before that? — illusion
– illusion, Commented Dec 2, 2020 at 20:24
No, first i call gradiantDescent and it doesn't work as i expected, so i call cost function separately to see if it work and it work correctly as it shown before. @illusion — Atef Magdy
– Atef Magdy, Commented Dec 2, 2020 at 20:32
In your code only theta[0] is getting updated. Shouldn't it run for all the thetas in theta array? — illusion
– illusion, Commented Dec 2, 2020 at 20:48
Or rather you are updating only one weight. Shouldn't you update all of them? — illusion
– illusion, Commented Dec 2, 2020 at 20:52
It's already run for all thetas. as the RHS in (theta = ...) is an array with shape(5,) so the theta array updated in every iterate, and i update theta[0] separately because it should take another value not as the other indices — Atef Magdy
– Atef Magdy, Commented Dec 2, 2020 at 20:54

illusion · Accepted Answer · 2020-12-03 07:03:02Z

1

Try this in your gradiantDescent function:

for _ in range(num_itr):
    theta = theta - (alpha / m) * np.dot(X.T, (np.dot(X, theta) - y))
    J_history.append(costFunction(theta, X, y))
return theta, J_history

You get a nan value because some calculations are going wrong...

answered Dec 3, 2020 at 7:03

illusion

1,30111 silver badges22 bronze badges

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

1 Comment

Atef Magdy Over a year ago

you are right theta values are going wrong, you alerted me to this point, i tried to find the error and it was the minus operand '-'. I changed it to numpy.subtract() and it's work correctly, Thanks for your effort.

Atef Magdy · Accepted Answer · 2020-12-03 11:39:06Z

1

The minus Operand was the error in calculating theta, Should use numpy.subtract(arr1, arr2) Old Code:

theta -= (alpha/m) * (np.dot(X.T , (sigmoid(np.dot(X,theta))-y)))

New:

np.subtract( theta ,(alpha/m) * (np.dot(X.T , (sigmoid(np.dot(X,theta))-y))) )

answered Dec 3, 2020 at 11:39

Atef Magdy

1229 bronze badges

Collectives™ on Stack Overflow

Python function behaves differently when i call it in another function loop

2 Answers 2

1 Comment

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

1 Comment

Comments

Your Answer

Sign up or log in

Post as a guest

Related