# Downsampling (machine learning)

> Overloaded term that can mean either of the following:[^1]

Overloaded term that can mean either of the following:[^1]

- Reducing the amount of information in a [feature](https://wiki.g15e.com/pages/Feature%20(machine%20learning.txt)) in order to [train](https://wiki.g15e.com/pages/Training%20(machine%20learning.txt)) a model more efficiently. For example, before training an image recognition model, downsampling high-resolution images to a lower-resolution format.
- Training on a disproportionately low percentage of over-represented [class](https://wiki.g15e.com/pages/Class%20(machine%20learning.txt)) examples in order to improve model training on under-represented classes. For example, in a [class-imbalanced dataset](https://wiki.g15e.com/pages/Class-imbalanced%20dataset.txt), models tend to learn a lot about the [majority class](https://wiki.g15e.com/pages/Majority%20class%20(machine%20learning.txt)) and not enough about the [minority class](https://wiki.g15e.com/pages/Minority%20class%20(machine%20learning.txt)). Downsampling helps balance the amount of training on the majority and minority classes.

## Footnotes

[^1]: https://developers.google.com/machine-learning/glossary#downsampling