Visualizing Overfitting and Regularization

How to Use

• Adjust Model Complexity to see how higher-degree polynomials fit the data (and noise)
• Increase Regularization to smooth the fit and reduce overfitting
• Click Reset Data to randomize the dataset
• Watch train/test errors update live

What You’re Seeing

• Blue circles: Training data points (what the model learns from)
• Red X markers: Test data points (unseen data used to evaluate generalization)
• Orange curve: Model fit curve
• Train/Test Error: Root mean squared error for each set

Key Observations

• High complexity fits noise (overfitting: low train error, high test error)
• Regularization smooths the fit, reducing overfitting
• Underfitting occurs at low complexity (high errors)
• The best fit balances bias and variance

Try different settings and see the bias-variance tradeoff in action!