What is a Feature Mask?



I was reading an article about a winning solution to the Criteo competition on Kaggle. There was also some talk of the third place solution. Here is the excerpt:

Highest scoring team using Vowpal Wabbit was Guocong Song for 3rd place. Method and code here. In short: Multiple models, polynomial learning and featuremasks.*

In this it mentions the use of Feature Mask or Featuremasks. Whatis meant by this? I am struggling to find good examples anywhere on the internet.

*emphasis my own.


@c3josh, As far as I understand, feature masks can be considered as feature selection. So what you do is mask the features which you consider as non-essential and don’t use them for training your model.

Read this article for a deeper understanding