What is the bias-variance tradeoff?
Updated May 15, 2026
Short answer
The bias-variance tradeoff describes the balance between model simplicity (bias) and complexity (variance) to minimize total error.
Deep explanation
The total error in machine learning can be decomposed into bias, variance, and irreducible error. Increasing model complexity reduces bias but increases variance. The optimal model finds a balance where both are minimized. This tradeoff is fundamental in model selection and regularization strategies.
Real-world example
Choosing between linear regression and random forest for house price prediction.
Common mistakes
- Thinking both bias and variance can be minimized simultaneously without tradeoffs.
Follow-up questions
- What is irreducible error?
- How does regularization affect the tradeoff?