STAR EMBEDDING PROJECTOR

Stellar Embeddings and Clustering Analysis

Interactive visualization of 160,000+ Milky Way stars using advanced dimensionality reduction and clustering techniques to reveal hidden patterns in high-dimensional astronomical data.

Interactive Demo

Explore the interactive 3D visualization of stellar embeddings above. Use mouse controls to rotate, zoom, and investigate different star clusters.

🔭Dataset Overview

Scale: 160,000+ stars from the Milky Way galaxy

Features: High-dimensional physical attributes including:

Temperature & Luminosity
Radius & Color Index
Spectral Type & Absolute Magnitude
Distance & Metallicity

Labels: Unique identifiers with stellar classifications

📉Dimensionality Reduction

✅ Principal Component Analysis (PCA)

Preserved linear variance, revealed wave-like patterns

✅ t-SNE

Captured local relationships and nonlinear similarities

✅ UMAP

Best preservation of local and global structure

📊Clustering Results & Key Findings

K-Means Clustering (10 Clusters)

White Dwarfs
Red Giants
Main Sequence Stars
Subgiants
Peculiar & Rare Classes

Scientific Insights

• Stars naturally group based on spectral and physical features
• UMAP revealed nonlinear continuity across star types
• Embedding methods enhanced interpretability of high-dimensional structure
• Clusters enable anomaly detection and candidate identification

💡Practical Applications

🔍

Classification

Astronomical classification and anomaly detection

📚

Education

Educational visualizations of star evolution

🤖

AI Training

Training data for AI-driven telescopic surveys

⚡

Simulation

Feature-rich input for astrophysics simulations