Assignment 1
Name: SAIKAT MOI
Roll Number: 20CS10050
# import all the necessary libraries here
import pandas as pd
import numpy as np
import random
import matplotlib.pyplot as plt
from sklearn.preprocessing import StandardScaler
from sklearn.model_selection import train_test_split
data = pd.read_csv('../../dataset/linear-regression.csv')
print(data)
fixed acidity volatile acidity citric acid residual sugar
chlorides \
0 7.4 0.700 0.00 1.9
0.076
1 7.8 0.880 0.00 2.6
0.098
2 7.8 0.760 0.04 2.3
0.092
3 11.2 0.280 0.56 1.9
0.075
4 7.4 0.700 0.00 1.9
0.076
... ... ... ... ...
...
1594 6.2 0.600 0.08 2.0
0.090
1595 5.9 0.550 0.10 2.2
0.062
1596 6.3 0.510 0.13 2.3
0.076
1597 5.9 0.645 0.12 2.0
0.075
1598 6.0 0.310 0.47 3.6
0.067
free sulfur dioxide total sulfur dioxide density pH
sulphates \
0 11.0 34.0 0.99780 3.51
0.56
1 25.0 67.0 0.99680 3.20
0.68
2 15.0 54.0 0.99700 3.26
0.65
3 17.0 60.0 0.99800 3.16
0.58
4 11.0 34.0 0.99780 3.51
0.56
... ... ... ... ...
...
1594 32.0 44.0 0.99490 3.45
0.58
1595 39.0 51.0 0.99512 3.52
0.76
1596 29.0 40.0 0.99574 3.42
0.75
1597 32.0 44.0 0.99547 3.57
0.71
1598 18.0 42.0 0.99549 3.39
0.66
alcohol quality
0 9.4 5
1 9.8 5
2 9.8 5
3 9.8 6
4 9.4 5
... ... ...
1594 10.5 5
1595 11.2 6
1596 11.0 6
1597 10.2 5
1598 11.0 6
[1599 rows x 12 columns]
X = data.drop(columns=['quality'])
y = data['quality']
# Feature scaling is a common preprocessing step in machine learning
that helps improve the performance and convergence
scaler = StandardScaler()
X_scaled = scaler.fit_transform(X)
X_train, X_temp, y_train, y_temp = train_test_split(X_scaled, y,
test_size=0.5, random_state=42)
X_val, X_test, y_val, y_test = train_test_split(X_temp, y_temp,
test_size=0.4, random_state=42)
print("Train set shapes - X:", X_train.shape, "y:", y_train.shape)
print("Validation set shapes - X:", X_val.shape, "y:", y_val.shape)
print("Test set shapes - X:", X_test.shape, "y:", y_test.shape)
Train set shapes - X: (799, 11) y: (799,)
Validation set shapes - X: (480, 11) y: (480,)
Test set shapes - X: (320, 11) y: (320,)
# concatenating the column vector of ones with the X_train matrix,
effectively adding a new column to the beginning of X_train
# bias term allows the model to learn an offset from the origin.
X_train_with_bias = np.c_[np.ones((X_train.shape[0], 1)), X_train]
# Normal Equation θ = (X^T X)^(-1) X^T y
theta = np.linalg.inv(X_train_with_bias.T @ X_train_with_bias) @
X_train_with_bias.T @ y_train
# Predict on the test set
X_test_with_bias = np.c_[np.ones((X_test.shape[0], 1)), X_test]
y_test_pred = X_test_with_bias @ theta
# Calculate R-squared on the test set
ssr = np.sum(( y_test-y_test_pred ) ** 2)
sst = np.sum((y_test - np.mean(y_test)) ** 2)
r2 = 1 - (ssr / sst)
# RMSE = sqrt((1 / n) * Σ(y_pred - y_actual)^2)
# Calculate RMSE on the test set
rmse = np.sqrt(np.mean((y_test - y_test_pred) ** 2))
print("Test Set - R-squared:", r2)
print("Test Set - Root Mean Squared Error (RMSE):", rmse)
Test Set - R-squared: 0.3967805084557149
Test Set - Root Mean Squared Error (RMSE): 0.6657115962476505
# the mean squared error loss between the predicted values and the
actual target values.
def compute_loss(X, y, theta):
y_pred = X @ theta
loss = np.mean((y_pred - y) ** 2)
return loss
# ∇(MSE) = -2 * (X^T @ (y - Xθ)) / N
def gradient_ascent(X, y, theta, learning_rate, num_iterations):
losses = []
for _ in range(num_iterations):
y_pred = X @ theta
gradient = -2 * (X.T @ (y - y_pred)) / len(y)
theta -= learning_rate * gradient
loss = compute_loss(X, y, theta)
losses.append(loss)
return theta, losses
learning_rates = [0.01, 0.001, 0.0001]
num_iterations = 50000
for lr in learning_rates:
theta_initial = np.zeros(X_train.shape[1] + 1) # +1 for the bias
term
X_train_with_bias = np.c_[np.ones((X_train.shape[0], 1)), X_train]
theta_optimized, losses = gradient_ascent(X_train_with_bias,
y_train, theta_initial, lr, num_iterations)
plt.plot(range(num_iterations), losses, label=f'lr = {lr}')
X_test_with_bias = np.c_[np.ones((X_test.shape[0], 1)), X_test]
y_test_pred = X_test_with_bias @ theta_optimized
# Calculate R-squared on the test set
ssr = np.sum(( y_test-y_test_pred ) ** 2)
sst = np.sum((y_test - np.mean(y_test)) ** 2)
r2 = 1-(ssr / sst)
# Calculate RMSE on the test set
rmse = np.sqrt(np.mean((y_test - y_test_pred) ** 2))
print(f"Learning Rate: {lr}")
print("Test Set - R-squared:", r2)
print("Test Set - Root Mean Squared Error (RMSE):", rmse)
print()
plt.xlabel('Iteration')
plt.ylabel('Loss')
plt.title('Loss Function vs. Iteration for Different Learning Rates')
plt.legend()
plt.show()
Learning Rate: 0.01
Test Set - R-squared: 0.3967805084557158
Test Set - Root Mean Squared Error (RMSE): 0.6657115962476501
Learning Rate: 0.001
Test Set - R-squared: 0.3967803979339991
Test Set - Root Mean Squared Error (RMSE): 0.6657116572333991
Learning Rate: 0.0001
Test Set - R-squared: 0.39639733535434873
Test Set - Root Mean Squared Error (RMSE): 0.6659229971434986
