Linear Regression

1. Regression Model
In statistics, linear regression is a linear approach to modelling the relationship between a dependent variable and one or more independent variables. Let x1x_1x1​
 be the independent variable and yyy
 be the dependent variable. We will define a linear relationship between these two variables as follows:
y=θ0+θ1x1y = \theta_0+\theta_1 x_1y=θ0​+θ1​x1​
2. Define Loss Function
We will use the Mean Squared Error function.
L=1n∑i=1n(ytrue−ypredicted)2L = \frac{1}{n}\sum_{i=1}^n (y_{true}-y_{predicted})^2L=n1​i=1∑n​(ytrue​−ypredicted​)2
3. Utilize the Gradient Descent Algorithm
You might know that the partial derivative of a function at its minimum value is equal to 0. So gradient descent basically uses this concept to estimate the parameters or weights of our model by minimizing the loss function.
1.
Initialize the weights, θ0=0\theta_0 = 0θ0​=0
and θ1=0\theta_1 =0θ1​=0
 
2.
Calculate the partial derivatives w.r.t. to θ0\theta_0θ0​
and θ1\theta_1θ1​
 
dθ0=−2n∑i=1n(yi−yiˉ)dθ1=−2n∑i=1n(yi−yiˉ)×xid_{\theta_0} = -\frac{2}{n} \sum_{i=1}^n(y_i - \bar{y_i}) \\   d_{\theta_1} = -\frac{2}{n} \sum_{i=1}^n(y_i - \bar{y_i}) \times x_idθ0​​=−n2​∑i=1n​(yi​−yi​ˉ​)dθ1​​=−n2​∑i=1n​(yi​−yi​ˉ​)×xi​
 
3.
Update the weights
θ0=θ0−l×dθ0θ1=θ1−l×dθ1\theta_0 = \theta_0 - l \times d_{\theta_0}  \\ \theta_1 = \theta_1 - l \times d_{\theta_1}θ0​=θ0​−l×dθ0​​θ1​=θ1​−l×dθ1​​
 
Python Implementation
1
# Importing libraries
2
import numpy as np
3
import pandas as pd
4
from sklearn.model_selection import train_test_split
5
​
6
# Preparing the dataset
7
data = pd.DataFrame({'feature' : [1,2,3,4,5,6,7,8,9,10,11,12,13,14,15], 'label' : [2,4,6,8,10,12,14,16,18,20,22,24,26,28,30]})
8
# Divide the data to training set and test set
9
X_train, X_test, y_train, y_test = train_test_split(data['feature'], data['label'], test_size=0.30)
10
​
11
# Method to make predictions
12
def predict(X, theta0, theta1):
13
    # Here the predict function is: theta0+theta1*x
14
    return np.array([(theta0 + theta1*x) for x in X])
15
​
16
def linear_regression(X, Y):
17
    # Initializing variables
18
    theta0 = 0
19
    theta1 = 0
20
    learning_rate = 0.001
21
    epochs = 300
22
    n = len(X)
23
​
24
    # Training iteration
25
    for epoch in range(epochs):
26
        y_pred = predict(X, theta0, theta1)
27
​
28
        ## Here the loss function is: 1/n*sum(y-y_pred)^2 a.k.a mean squared error (mse)
29
        # Derivative of loss w.r.t. theta0
30
        theta0_d = -(2/n) * sum(Y-y_pred)
31
        # Derivative of loss w.r.t. theta1
32
        theta1_d = -(2/n) * sum(X*(Y-y_pred))
33
​
34
        theta0 = theta0 - learning_rate * theta0_d
35
        theta1 = theta1 - learning_rate * theta1_d   
36
​
37
    return theta0, theta1
38
​
39
# Training the model
40
theta0, theta1 = linear_regression(X_train, y_train)   
41
​
42
# Making predictions
43
y_pred = predict(X_test, theta0, theta1)
44
​
45
# Evaluating the model
46
print(list(y_test))
47
print(y_pred)
Copied!
​

Machine Learning ↓↑ - Previous

L1 and L2 Loss Function

Next - Machine Learning ↓↑

Logistic Regression

Last modified 9mo ago

Copy link

Contents

1. Regression Model

2. Define Loss Function

3. Utilize the Gradient Descent Algorithm

Python Implementation