UNIVERSITY OF ENGINEERING AND TECHNOLOGY, TAXILA
LAB REPORT # 02
MACHINE LEARNING (LAB)
NAME : Muhammad Islam ul haq
ROLL NO. : 21-CP-81
SECTION : Alpha
TASKS
Task 1:
Statement:
Download the folder from Teams Lab 03 Pre_tasks and upload the notebook in google colab and upload
the required csv your task is to run and analyze the code write your understanding.
Answer:
After uploading and analyzing the code we came to know to know that Linear Regression is good untill
line for data points is straightforward when data points start to come in random location than it is
difficult to predict value so for that purpose linear regression is not good.
Task 2:
Statement:
Download the 3_gradient_descent zip file from Teams and Extract it find the files and Exercise folder run
the code and get your understanding ?
Answer:
This code is an implementation of the gradient descent algorithm for simple linear regression. The goal of
the algorithm is to find the best-fitting line (linear relationship) between a single independent variable (x)
and a dependent variable (y). The code iteratively updates the slope (m) and y-intercept (b) of the line to
minimize the mean squared error between the predicted and actual values.
Task 3:
Multivariate linear regression.
Statement:
Modify the Example 1 code to handle multivariate linear regression. Update the synthetic dataset
to include multiple features.
Code:
import numpy as np
import pandas as pd
def gradient_descent(X, y, learning_rate, num_iterations):
# Initialize coefficients
theta = np.zeros(X.shape[1])
m = len(y)
for iteration in range(num_iterations):
# Calculate predictions
predictions = np.dot(X, theta)
# Calculate errors
errors = predictions - y
# Update coefficients
gradient = np.dot(X.T, errors) / m
theta -= learning_rate * gradient
# Calculate and print the cost
cost = np.sum(errors**2) / (2 * m)
print(f"Iteration {iteration + 1}/{num_iterations}, Cost: {cost}")
return theta
# Load data from CSV file
data = pd.read_csv('/canada_per_capita_income.csv')
# Extract 'year' and 'per_capita_income' columns
X = data[['year']].values
y = data['per_capita_income'].values
# Add a column of ones to X for the intercept term
X_b = np.c_[np.ones((X.shape[0], 1)), X]
# Apply gradient descent
learning_rate = 0.01
num_iterations = 10
theta = gradient_descent(X_b, y, learning_rate, num_iterations)
print("Final Coefficients:", theta)
OUTPUT:
Task 4:
Statement:
Modify the Example 1 code to implement stochastic gradient descent. Update the algorithm to update the
coefficients based on a single randomly chosen instance at each iteration.
Code:
import numpy as np
def stochastic_gradient_descent(X, y, learning_rate, num_iterations):
# Initialize coefficients
theta = np.zeros(X.shape[1])
m = len(y)
for iteration in range(num_iterations):
# Randomly shuffle the data indices
indices = np.random.permutation(m)
for i in indices:
# Select a random instance
X_i = X[i:i+1]
y_i = y[i:i+1]
# Calculate prediction
prediction = np.dot(X_i, theta)
# Calculate error
error = prediction - y_i
# Update coefficients based on the single instance
gradient = np.dot(X_i.T, error)
theta -= learning_rate * gradient
# Calculate and print the cost for the entire dataset
predictions = np.dot(X, theta)
errors = predictions - y
cost = np.sum(errors**2) / (2 * m)
print(f"Iteration {iteration + 1}/{num_iterations}, Cost: {cost}")
return theta
# Usage example
X = np.random.rand(100, 1) # Example feature
y = 2 * X + 1 + 0.1 * np.random.randn(100, 1) # Example linear relationship with noise
# Add a column of ones to X for the intercept term
X_b = np.c_[np.ones((100, 1)), X]
# Apply stochastic gradient descent
learning_rate = 0.01
num_iterations = 100
theta = stochastic_gradient_descent(X_b, y, learning_rate, num_iterations)
print("Final Coefficients:", theta)
OUTPUT:
------------------------------------------------------------------------------
