Only pull unique values in generate_vocab #502
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This patch makes generate_vocab only consider unique values when creating quantile buckets for the individual features. There is no point in mapping multiple of the same value to different values post normalization/bucketization. This only serves to decrease dynamic range. Only looking at unique values preserves the property of increasing dynamic range around areas that have lots of slightly different values. This is especially useful when generating vocab for features that are oftentimes zero or one. In some cases I was seeing <50/1000 actual values with the rest being zeros or ones, which results in models that train poorly.
|
Running some BC hparam explorations right now with one of the parameters I'm exploring being vocab. I'm ablating against the current vocab in tree, new vocab without this patch, and new vocab with this patch. I'll try and post the results here once I have them in a couple hours. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This patch makes generate_vocab only consider unique values when creating quantile buckets for the individual features. There is no point in mapping multiple of the same value to different values post normalization/bucketization. This only serves to decrease dynamic range. Only looking at unique values preserves the property of increasing dynamic range around areas that have lots of slightly different values. This is especially useful when generating vocab for features that are oftentimes zero or one. In some cases I was seeing <50/1000 actual values with the rest being zeros or ones, which results in models that train poorly.