🎭

Overfitting Theater

From: The Overfitting Trap•View in lesson →

💡

When is "perfect" actually bad? In Overfitting Theater, you'll witness a model achieve 100% accuracy on training data while completely failing on new examples. This visualization reveals one of the most important concepts in machine learning: the danger of memorization over generalization.

🎯
Sweet Spot!
Simple (Straight Line)Complex (Very Wiggly)
1-2
Underfitting
3-4
Just Right
8+
Overfitting
Training Error
0.000
Test Error
0.000

Observations

📘

Underfitting: Model is too simple. It can't capture the pattern, so both training AND test errors are high.

Understanding the Theater

  • 🎭 The Drama: As complexity increases, training error always decreases, but test error eventually increases
  • 🔵 Blue dots: Training data the model learns from
  • 🔴 Red dots: Test data the model has never seen
  • 📈 The curve: What the model learned. Watch it get increasingly wiggly!
  • 🎯 Goal: Find the complexity that minimizes TEST error, not training error

Experiment as much as you like; there is no progress to lose.