Section 1 — What you will learn
✓ By the end you will understand
- Node = a question the AI asks.
- Split = “If feature ≤ threshold → left, else → right”.
- Impurity (Gini / Entropy) = “how mixed” the node is.
- Information Gain = “how much cleaner the children become”.
- Depth = how many questions it can ask (too deep → overfit).
- Pruning = cutting weak branches to generalize better.
How to play:
1) Press Step Grow to build 1 split at a time (best for kids).
2) Press Run Packets to see data travel from root → leaf (decision path).
3) Use Break It 💥 to add noise and watch the tree overfit.
1) Press Step Grow to build 1 split at a time (best for kids).
2) Press Run Packets to see data travel from root → leaf (decision path).
3) Use Break It 💥 to add noise and watch the tree overfit.
Section 2 — Controls
Story helps kids remember what “class” means.
Decision trees draw only vertical/horizontal splits (axis-aligned), so curves become stair-steps.
Both measure “mixing”. Different math, same idea.
More depth = more questions. Too much depth = overfitting monster.
A node splits only if it has at least this many samples.
Higher = cuts more weak branches (simpler tree).
Noise = wrong labels in real life (bad sensors / mistakes / messy data).
This controls node spacing and edge lengths (prevents “chipke-chipke” look).
Section 3 — Live stats
Nodes
—
Leaves
—
Train Accuracy
—
Test Accuracy
—
Overfit Gap
—
Noise Level
—
Live Chart
Train/Test accuracy (updates as tree grows)
Section 4 — Live math (kid-friendly)
Press Step Grow to see the math for the next best split.
🧠 Kid Coach
A decision tree learns by asking Yes/No questions. It keeps splitting until nodes become “clean”.
Try: increase depth → watch more branches grow.