Section 1 — What you will learn
✓ By the end you will understand
- The line formula: ŷ = m·x + b (m = slope, b = bias)
- Error stick: e = y − ŷ (how far point is from line)
- Score (loss): MSE = (1/n) Σ (y − ŷ)²
- Training means: change m and b to make MSE smaller
- Gradient descent idea: take small steps downhill
Section 2 — Controls
Slope = tilt. More tilt → line becomes steeper.
Bias = lift up/down. It moves the line without tilting it.
Slow speed = best for kids (you can SEE m & b changing).
Small LR = safe & smooth. Big LR = jumpy.
Auto Train shows the model adjusting m & b internally. “1 Train Step” is super kid-friendly.
Section 3 — Live stats
Current MSE
0.0000
Best MSE (so far)
0.0000
MSE health bar (lower is better)
Training step count
0
Model’s “inside brain” (gradients)
∂MSE/∂m = 0.000
∂MSE/∂b = 0.000
∂MSE/∂b = 0.000
Sign tells direction: if ∂MSE/∂m is positive, decrease m.
Section 4 — Formulas (LIVE)
🧠 Kid Coach
Section 5 — Error Sticks Playground
data points
prediction line
error sticks
“downhill direction”
best line memory
Live explanation
Move m & b. Auto Train will show HOW the model changes them.