Machine Learning Gradient descent python implementation

Question

Problem

I have written this code, but this is giving errors:

RuntimeWarning: overflow encountered in multiply
t2_temp = sum(x*(y_temp - y))
RuntimeWarning: overflow encountered in double_scalars
t1_temp =  sum(y_temp - y)

Should I use feature scaling or is there something wrong in my code?

Code

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

def gradient_descent(x,y,t1,t2,repeat,alpha):
    n = x.size
    for i in range(repeat):
        y_temp = x*t2 + t1
        t1_temp =  sum(y_temp - y)
        t2_temp = sum(x*(y_temp - y))
        t1 = t1 - alpha * (t1_temp/n)
        t2 = t2 - alpha * (t2_temp/n)

    return [t1,t2]


d = pd.read_csv("train.csv")

x = d['GrLivArea']
y = d['SalePrice']

x = (np.array(x.values))
y = (np.array(y.values))


alpha = 0.047
repeat = 3000

theta = [1.23154644,1.654132398]

tt = gradient_descent(x,y,theta[0],theta[1],repeat,alpha)

print("FINISH")

Try scaling your values in the columns by a factor, and run it. Most likely the column values are too big, causing overflow. — hlkstuv_23900
– hlkstuv_23900, Commented Mar 22, 2019 at 4:36

gmds · Accepted Answer · 2019-03-22 06:01:55Z

1

Your code is fine; it is x and y that are too large. You can perform standardisation or some other form of scaling.

answered Mar 22, 2019 at 6:01

gmds

20k4 gold badges37 silver badges65 bronze badges

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

6 Comments

K. Bakshi Over a year ago

I am applying but for testing data i am getting wrong output. Feature scaling will give the value of theta1 and theta2 which we can use as it is in our equation or testing data should also be processed further ?

K. Bakshi Over a year ago

After getting value of theta1 and theta2 i am putting it into equation and i am checking my test data , that is basically plot size , so the output which should be price is also scaled down.

gmds Over a year ago

If you have not scaled your y values, then the scaling of x should only affect the coefficients, not the results. Have your coefficients converged? (check if the gradients are near 0)

K. Bakshi Over a year ago

Yes i have scaled both x and y and my coefficients are also converging both are near 0.

gmds Over a year ago

@K.Bakshi well, in that case you would need to scale y up again.

|

Collectives™ on Stack Overflow

Machine Learning Gradient descent python implementation

Problem

Code

1 Answer 1

6 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

Problem

Code

1 Answer 1

6 Comments

Your Answer

Sign up or log in

Post as a guest

Related