ML Math Lab — PCA: The Data Squisher
Center → spin the laser → lock PC1 → squish data → keep maximum variance (Level 2: 2D + 3D Collapse)
Phase: Ready
Hint: press Start
Story: Generic
Section 1 — What you will learn
✓ By the end you will understand
  • Variance: how “spread out” your data is.
  • PC1: the direction where data spreads the most (maximum variance).
  • Projection: dropping shadows of points onto the PC line.
  • Information retained: how much variance we kept after squishing.
  • Why PCA exists: compress data while keeping the main structure.
How to play (super simple):
1) Press Start → watch the 3 cinematic phases.
2) Move Manual Angle to feel “wrong angle vs best angle”.
3) Try Break it 💥 → see PCA fail when the data is too symmetric/noisy.
4) Press 3D Mode → watch a 3D cloud collapse into 2D.
Section 2 — Controls
Story helps kids remember what “information retained” means.
Manual override: turn it ON to feel why PCA chooses the best angle.

Section 3 — Live stats
Information retained
Information lost
PC1 power (eigenvalue)
PC2 power (eigenvalue)
Reconstruction error
Phase progress
Live chart
Retained variance & error trend (updates live)
Scree plot
How much each component explains

Section 4 — Live math (kid-simple)
Loading…
🧠 Kid Coach
Loading…
Tip: PCA chooses the angle with maximum spread.