PCA — Principal Component Analysis
Interactive Demo via SVD · drag to rotate · scroll to zoom
Hanbyul Joo
| Visual Computing Lab, Seoul National University |
More Demos ↗
3D Data Space · drag: rotate · wheel: zoom
Step 0 – Original 3D Data
2D Projection: PC1 × PC2
1D Projection: PC1 axis only
Data Shape
Scale λ₁
3.2
Scale λ₂
1.4
Scale λ₃
0.5
Rot X
30°
Rot Y
45°
Rot Z
15°
Samples
250
Regenerate Data
Explained Variance (Eigenvalues)
PC1
—
PC2
—
PC3
—
λ₁
=
—
λ₂
=
—
λ₃
=
—
Cumulative (PC1+2):
—
%
Σ = V · diag(λ) · Vᵀ
Columns of V = principal axes
Covariance Matrix Σ
—
—
—
—
—
—
—
—
—
Visualization Options
Show PC axes
Covariance ellipsoid
Projection plane (PC1-PC2)
Color by PC1 score
Dimensionality Reduction
3D Original
→ 2D
→ 1D
Recon. Error
Variance decreases along PC1 → PC2 → PC3.
PC axes are mutually orthogonal.