Bias-Variance Tradeoff

noun

say it: “BY-us VAIR-ee-unss TRAYD-awf” /ˈbaɪəs ˈvɛriəns ˈtreɪdˌɔf/

The balance in machine learning between a model too simple to capture a pattern and one so complex that it also captures random noise.

What is the bias-variance tradeoff?

The bias-variance tradeoff is the balance between a model too simple to catch the pattern (high bias: underfitting) and one so flexible it learns each example's noise (high variance: overfitting). Making a model more flexible lowers one and raises the other, so the best setting sits in between.

In a sentence

  • “The validation curve made the bias-variance tradeoff easy to see.”
  • “Picking the tree's depth is the bias-variance tradeoff in action: too shallow misses the pattern, too deep memorizes noise.”

Level: Freelancer · More AI Basics terms at this level

Updated