Dimensionality Reduction: PCA vs t-SNE vs UMAP

Overview

This project compares three dimensionality reduction techniques—PCA, t-SNE, and UMAP—on high-dimensional datasets to evaluate how well each method preserves structure and class separability in low-dimensional embeddings.

Methods

• PCA (Principal Component Analysis)
• t-SNE (t-Distributed Stochastic Neighbor Embedding)
• UMAP (Uniform Manifold Approximation and Projection)

Data

Scikit-learn benchmark datasets:

• Digits dataset (1797 samples, 10 classes)
• Breast Cancer dataset (569 samples, binary classification)

Tools

Python, NumPy, scikit-learn, UMAP, Matplotlib, Seaborn