Technical interview guide

Bias-Variance Tradeoff

Why model error splits into bias and variance, and why reducing one often increases the other.

Read: 24 min
Practice MCQs: 25
Interview QA: 25
Edition: v5
Editorial status: Reviewed
Relevant for: AI EngineerMLOps EngineerMachine Learning Engineer

Scope: scikit-learn stable model-selection and ensemble guidance and Google ML Crash Course overfitting guidance accessed 2026-08-31..

Interview QA

Treat each question like a live interview question: answer out loud first (structure, assumptions, tradeoffs), then open the model answer to spot gaps and rehearse a tighter follow-up.

Curated: 2026-10-05 · Written: 2026-10-05 · Reviewed: 2026-10-03

QA-1

Explain bias, variance, and irreducible noise.

QA-2

How do underfitting and overfitting appear in training and validation results?

QA-3

How would you diagnose a large train-validation gap?

QA-4

How would you diagnose poor training and validation performance?

QA-5

How do you construct and interpret a learning curve?

QA-6

How do you construct and interpret a validation curve?

QA-7

When will more data reduce variance?

QA-8

When will more data fail to fix model error?

QA-9

How does regularization change the bias-variance tradeoff?

QA-10

Explain early stopping as a regularizer.

QA-11

How does bagging reduce variance, and when will it not help much?

QA-12

Compare bagging and boosting through bias and variance.

QA-13

Why does stacking require out-of-fold predictions?

QA-14

How do you select model complexity without overusing the test set?

QA-15

Explain selection bias from large hyperparameter searches.

QA-16

How would you use nested cross-validation in a small dataset?

QA-17

How do label noise and irreducible noise differ?

QA-18

Can a model have both high bias and high variance?

QA-19

How does feature engineering affect bias and variance?

QA-20

How does data augmentation affect bias and variance?

QA-21

How do you distinguish overfitting from distribution shift?

QA-22

How do you diagnose validation performance that is better than training performance?

QA-23

How do you choose between a simpler model and more regularization?

QA-24

How should bias and variance be monitored after deployment?

QA-25