Assignment 4: Unsupervised Learning - Dimensionality Reduction and Clustering

This repository contains the Jupyter Notebook for Assignment 4 of the Elements of Machine Learning course. The assignment focuses on unsupervised learning techniques, specifically clustering and dimensionality reduction.

Contents

1. Clustering

• Dataset: Digits dataset from scikit-learn (8x8 pixel images of handwritten digits).
• Objective: Cluster the data into groups without using the labels.
• Key Tasks:
  • Visualizing sample digits.
  • Using the k-means clustering algorithm to group the data.
  • Determining the optimal number of clusters using the elbow method.
  • Exploring clustering as a preprocessing step for semi-supervised learning.

2. Dimensionality Reduction (PCA)

• Objective: Apply Principal Component Analysis (PCA) to reduce the dimensionality of the data while retaining maximum variance.
• Key Tasks:
  • Perform PCA on the Digits dataset.
  • Train and evaluate a Logistic Regression model with different numbers of PCA components.
  • Identify the optimal number of PCA components for a balance between performance and complexity.

Dependencies

The following Python libraries are required to run the notebook:

• numpy
• matplotlib
• scikit-learn