Popular Courses
Linear Regression
Introduction
Decision Trees
Introduction
Neural Networks
Introduction

Sign in to Curvingo

or

New Account     Reset Password     

Testing Your AI
Introduction
Our Previous Approach
K-Fold Cross-Validation
Leave-One-Out Cross Validation (LOOCV)
See For Yourself
Which K Should You Pick?
Conclusion
Built with Mathigon

Share

Send us Feedback

Please let us know if you have any feedback and suggestions, or if you find any errors and bugs in our content.

Sorry, your message couldn’t be submitted. Please try again!

Thanks for your feedback!

Reset Progress

This will delete your progress and chat data for all chapters in this course, and cannot be undone!

Glossary

Select one of the keywords on the left…

Testing Your AILeave-One-Out Cross Validation (LOOCV)

Reading time: ~5 min

What happens if we push K-Fold Cross-Validation to its absolute, most extreme limit?

This extreme scenario is called Leave-One-Out Cross-Validation (LOOCV). It occurs when we set the number of chunks (k) to be exactly equal to the total number of individual data points in our entire dataset (n).

Instead of holding back a large exam chunk, your validation "exam" is literally just ! The AI trains on every single piece of data except for one left-out observation, tries to predict that missing piece, and then wipes its brain to start over.

Terrifying Computing Costs

Because the AI gets to train on n - 1 data points, it is seeing nearly 100% of the entire dataset during every single training run. This makes LOOCV incredibly accurate and practically eliminates variance... but it comes at a terrifying cost.

If you have a dataset with 100,000 photos... LOOCV requires you to train 100,000 completely separate AI models from scratch!

Because of this astronomical computing cost, you will almost never see data scientists actually use LOOCV in the real world unless they are working with tiny datasets or very specific mathematical shortcut models. Instead, most standard machine learning projects stick to much safer, more conservative values like k = 5 or k = 10.

Sina