Essay

When Additional Data Helps Reduce Variance

Question: When is collecting more training data a practical way to reduce high variance, and what usually happens to bias when you do this?

Sample answer: Adding more training examples can help with high variance when you can obtain a substantially larger dataset and you have enough computing resources to train on it. This approach usually lowers variance, but it does not meaningfully change bias.

Key points:

  • More data is only practical when a much larger dataset is available.
  • Sufficient computing resources are needed to use the extra data.
  • More training data usually reduces variance.
  • More training data usually leaves bias unchanged.

Rubric: The response must state that adding data is useful only when significantly more data and enough computation are available. It must also explain that this typically reduces variance without changing bias.

0

1

Updated 2026-08-12

Contributors are:

Who are from:

Tags

Machine Learning

Deep Learning

Supervised Learning

Dive into Deep Learning @ D2L

Data Science

Machine Learning Strategy

Machine Learning Yearning @ DeepLearning.AI