Section 1 — What you will learn
✓ By the end you will understand
- K-Fold: every chunk becomes the exam once.
- One lucky split can lie: K-Fold checks consistency.
- Mean accuracy: average score across folds.
- Std deviation (“fear meter”): high std = unreliable model.
- Fraud demo: looks good once, fails across folds.
How to learn (step-by-step):
1) Press Single Split → see “lucky” score.
2) Press Start → watch K-Fold rotation & scoring.
3) Use Break it 💥 → noise ↑ → std deviation ↑ (unreliable).
4) Change K & Seed → see consistency change.
1) Press Single Split → see “lucky” score.
2) Press Start → watch K-Fold rotation & scoring.
3) Use Break it 💥 → noise ↑ → std deviation ↑ (unreliable).
4) Change K & Seed → see consistency change.
Section 2 — Controls
More folds = more exams = better reliability check.
Different models can be stable or unstable across folds.
Change seed to see how luck can fake a single split.
Story helps kids remember what the score means.
Fraud demo = noisy world.
Section 3 — Live stats (Scoreboard)
Fold score (current)
—
All fold scores
—
Mean accuracy
—
Std deviation (fear)
—
Reliability
—
Single-split accuracy
—
Live Chart
Fold accuracy over time
Section 4 — Live math (super simple)
Loading…
🧠 Kid Coach
Loading…
Tip: low std deviation = reliable model.