Short Answer

Subsets of Error Sources in Machine Learning Algorithms

Question: In machine learning development, must avoidable bias, variance, and data mismatch always occur together, or can an algorithm suffer from only a specific subset of these three problems?

Sample answer: An algorithm does not need to suffer from all three problems together. It is possible for an algorithm to suffer from any subset of high avoidable bias, high variance, and data mismatch, meaning it can exhibit any combination of these errors (e.g., just variance and data mismatch, or all three, or only one) depending on the model and data.

Key points:

  • Avoidable bias, variance, and data mismatch do not have to occur together.
  • An algorithm can suffer from any subset of high avoidable bias, high variance, and data mismatch.

Rubric: The answer should state that errors do not have to occur together and that a model can suffer from any subset or combination of avoidable bias, variance, and data mismatch.

0

1

Updated 2026-05-26

Contributors are:

Who are from:

Tags

Machine Learning

Deep Learning

Supervised Learning

Dive into Deep Learning @ D2L

Data Science

Machine Learning Strategy

Related