Python function returning wrong value

Question

I have a function (gradient descent) in python that return me some values:

import pandas as pd
import numpy as np
import matplotlib.pyplot as plt


def read_data(file):
    df = pd.read_excel(file)
    x_data= np.array(df['X_axis']) 
    y_data = np.array(df['Y_axis']) 

    return x_data,y_data 


x_data ,y_data = read_data('path\file')

alpha= 10**-8

#Auxiliaries Functions
...

here is the function:

def gradient(x_axis,y_axis):

    current_iteration = 0
    iterations_number= 100

    #Initial values
    #I guess that the problem is here. When I print those initial 
    #values,inside the function, they are different from what they should be   


    A1_inicial = y_axis[0] 
    A2_inicial = y_axis[-1] 
    x0_inicial = np.mean(x_axis) 
    dx_inicial = (np.std(x_axis))   


    while current_iteration < iterations_number:
        sum_A1 = 0
        sum_A2 = 0
        sum_dx = 0
        sum_x0 = 0


       for x_values,y_values  in zip(x_axis, y_axis):
            sum_A1 += derivada_A1(A1_inicial,A2_inicial,x0_inicial,dx_inicial,x_values,y_values  )
            sum_A2 += derivada_A2(A1_inicial,A2_inicial,x0_inicial,dx_inicial,x_values,y_values  )
            sum_dx += derivada_dx(A1_inicial,A2_inicial,x0_inicial,dx_inicial,x_values,y_values  )
            sum_x0 += derivada_x0(A1_inicial,A2_inicial,x0_inicial,dx_inicial,x_values,y_values  )



            A1_inicial = A1_inicial - (alpha* sum_A1 )
            A2_inicial = A2_inicial - (alpha* sum_A2 )
            x0_inicial = x0_inicial - (alpha* sum_x0 )
            dx_inicial = dx_inicial - (alpha* sum_dx )



        return A1_inicial,A2_inicial,x0_inicial,dx_inicial
        current_iteration+=1

So, the problem is that, whenever I use this function, my values that are being returned, are wrong. I know that they are wrong because I have the real values to compare to.

Here are how I am using the function:

new_y = []

A1_inicial,A2_inicial,x0_inicial,dx_inicial = gradient(x_axis,y_axis)

for x in x_axis:
    new_y.append(A2_inicial + ((A1_inicial - A2_inicial) /(1+np.exp((x- x0_inicial)/dx_inicial))))

print("A1: {}".format(A1_inicial))
print("A2: {}".format(A2_inicial))
print("X0: {}".format(x0_inicial))
print("DX: {}".format(dx_inicial))

And I know that if I undo this function, like this:

current_iteration = 0
iterations_number= 100


A1_inicial = y_axis[0] 
A2_inicial = y_axis[-1] 
x0_inicial = np.mean(x_axis) 
dx_inicial = (np.std(x_axis))   


while current_iteration < iterations_number:
    sum_A1 = 0
    sum_A2 = 0
    sum_dx = 0
    sum_x0 = 0


    for x_values,y_values in zip(x_axis, y_axis):
        sum_A1 += derivada_A1(A1_inicial,A2_inicial,x0_inicial,dx_inicial,x_values,y_values  )
        sum_A2 += derivada_A2(A1_inicial,A2_inicial,x0_inicial,dx_inicial,x_values,y_values  )
        sum_dx += derivada_dx(A1_inicial,A2_inicial,x0_inicial,dx_inicial,x_values,y_values  )
        sum_x0 += derivada_x0(A1_inicial,A2_inicial,x0_inicial,dx_inicial,x_values,y_values  )

    A1_inicial = A1_inicial - (alpha* sum_A1 )
    A2_inicial = A2_inicial - (alpha* sum_A2 )
    x0_inicial = x0_inicial - (alpha* sum_x0 )
    dx_inicial = dx_inicial - (alpha* sum_dx )

And then:

new_y = []


for x in x_axis:
    new_y.append(A2_inicial + ((A1_inicial - A2_inicial) /(1+np.exp((x- x0_inicial)/dx_inicial))))

print("A1: {}".format(A1_inicial))
print("A2: {}".format(A2_inicial))
print("X0: {}".format(x0_inicial))
print("DX: {}".format(dx_inicial))

It works, giving me the right values, but I don't know what is wrong with the function and whyt it does not work.

Thanks for the help!

Answer 1

In your gradient function, you have the return statement before you are incrementing your current_iteration variable. So it's never getting incremented. The function terminates at the point of return.

Add the current_iteration incrementor before return statement, and add the return statement after your while loop ends.