ML Math Lab — Model Training Factory™
Train the Brain → Control Chaos → Achieve Generalization (with a Boss Fight)
Mode: Explore
Goal: generalize
Score: 0
Section 1 — What you will learn
By the end you will understand
  • Underfitting (High Bias): the brain is too stiff → it can’t bend to the real rule.
  • Overfitting (High Variance): the brain gets too wiggly → it memorizes noise.
  • Generalization: the brain learns the pattern → works on new data too.
  • Noise (Chaos): real life is messy → data shakes and becomes fuzzy.
  • Regularization (λ): a stability shield → prevents “spaghetti thinking”.
Section 2 — Controls
Challenge adds scoring + win banner + Boss Chaos test.
Degree 1 is a straight stick. Degree 12 can become spaghetti.
Higher λ reduces wiggles (variance), but can increase bias.
Noise is visual: capsules shake + truth gate becomes stormy.
More points → brain becomes more reliable.
How to play: Click in the Training Chamber → you send a probe pulse at x. The brain predicts y. You see error beams (how wrong it is).
Yellow capsules = training. Cyan capsules = reality test (hidden until Reveal).

Section 3 — Live stats
Train MSE
Test MSE
Hidden
Bias Meter (stiffness)
Variance Meter (wiggles)
Generalization Gap
Hits

Section 4 — Math (very easy)
What the brain is doing Hover terms • Click to probe
Loading…
🧠 Kid Coach
Loading…
Tip: Reveal Reality = real exam.
Section 5 — The Training Factory
training capsules reality capsules (test) brain curve (prediction engine) error beam (miss) stability shield glow hidden truth guide
Live explanation
Click in the chamber → send a probe pulse. Reveal Reality when you’re confident.
Game HUD
Explore: tune the brain to be smart, not crazy.
Overfit Meter (gap)
Reveal Reality to measure overfitting.
Wiggle Meter (variance)
High wiggle = memorizing noise.
Stiffness Meter (bias)
Too stiff = can’t bend toward truth.
Quick actions
Challenge win = low Train + low Test + small gap + not too wiggly. Then pass Boss Chaos.