EngineersofAI LogoCoursesBlogResearch LabAI LettersThe LabCode BankInteractive 3DKodrEarnestJobs
Skip to main content
Interactive 3D/Bias-Variance Tradeoff
Orange line = current degree · Grey dashed = true function
Polynomial Degree
degree = 4Good fit
110
Data
Errors (MSE)
Training:-
Test:-
Zones
Underfitting (degree 1-2): model too simple - misses the pattern entirely.
Good fit (degree 3-5): captures the signal, generalises to new data.
Overfitting (degree 7+): memorises training noise, fails on test set.
Key Insight
Low degree → high bias (underfits, misses the pattern). High degree → high variance (overfits, memorises noise). The optimal degree minimises test error - not training error.
Try: Click "New Data" several times with degree 9 - watch the curve wildly change. Then switch to degree 3 - it stays stable.

Bias-Variance Tradeoff - Interactive Visualization

The bias-variance tradeoff is the fundamental tension in machine learning: simple models underfit (high bias, miss the pattern), complex models overfit (high variance, memorize noise). This visualization fits polynomial regression models of degree 1–10 to noisy data, showing training error, test error, and the classic U-shaped generalization curve.

  • Degree 1-2: high bias - straight line misses the curve entirely
  • Degree 3-5: the sweet spot - captures the true function, generalizes well
  • Degree 7+: high variance - wiggles through every training point but fails on test data
  • Error curve chart shows training vs test MSE for all degrees simultaneously
  • Click "New Data" to see how the optimal degree is consistent across different noise realizations
  • Direct foundation for regularization (L1/L2), dropout, early stopping, and cross-validation

Part of the EngineersOfAI Interactive 3D - free interactive visualizations covering every major concept in machine learning and AI engineering. Hover any element for a plain-English explanation. No code required.