🎨 t-SNE Visualization

t-SNE (t-Distributed Stochastic Neighbor Embedding)

Purpose: Reduce high-dimensional data to 2D/3D for visualization

How: Preserves local structure - similar points stay close together

vs PCA: t-SNE captures non-linear relationships, PCA is linear

Perplexity: Balances local vs global structure (typical: 5-50)

Uses: Visualize embeddings • Explore clusters • Feature analysis • Quality check