feat: Add classification functions #11792
Open
+1,249
−7
Conversation
This file contains 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
facebook-github-bot
added
the
CLA Signed
|
This pull request was exported from Phabricator. Differential Revision: D66684198 |
✅ Deploy Preview for meta-velox canceled.
|
yuandagits
added a commit
to yuandagits/velox
that referenced
this pull request
Dec 9, 2024
Summary: Add the classification functions from presto into velox: https://prestodb.io/docs/current/functions/aggregate.html#classification-metrics-aggregate-functions Classification functions all use `FixedDoubleHistogram`, which is a data structure to represent the bucket of weights. The index of the bucket for the histogram is evenly distributed between the min and value values. For all of the classification functions, the only difference is the extraction phase. All other steps will be the same. At a high level: - addRawInput will add a value into either the true or false weight bucket. The bucket to add the value to will depend on the prediction value. The prediction value is linearly mapped into a bucket based on (min, max and bucketCount) by normalizing the prediction between min and max. - The schema of the intermediate states is [version header][bucket count][min][max][weights] Differential Revision: D66684198
yuandagits
force-pushed
the
export-D66684198
branch
from
e939248 to
d2b00ed
Compare
|
This pull request was exported from Phabricator. Differential Revision: D66684198 |
yuandagits
changed the title
Yuhta
reviewed
Dec 9, 2024
| in.copyTo(&min, 1); | ||
| in.copyTo(&max, 1); | ||
|
|
||
| auto ret = FixedDoubleHistogram(bucketCount, min, max, allocator); |
There was a problem hiding this comment.
Can we avoid memory allocation for the buckets? Just merge with a view on the deserialized bytes (be careful about the alignment though).
| /// std::vector<double>::max_size(), which may be less than 2^63 depending. To | ||
| /// account for this, we have two buckets which may be used to store the | ||
| /// weights with each bucket being at most kMaxBucketCount in size. | ||
| static constexpr int64_t kMaxBucketCount = |
There was a problem hiding this comment.
In practice max_size is at least 2^60 on 64 bits system, I don't think any system can give that large contiguous memory in one go. So you don't need to split the weights into 2 arrays for the discrepancy between 2^60 vs 2^64 (you can put a VELOX_CHECK_LT(bucketCount, weights_.max_size()) in validateParameters if you are really concerned about this).
There was a problem hiding this comment.
Agreed, I added this because it was something the fuzzer caught
yuandagits
force-pushed
the
export-D66684198
branch
from
d2b00ed to
6bf72c8
Compare
yuandagits
added a commit
to yuandagits/velox
that referenced
this pull request
Dec 11, 2024
Summary: Add the classification functions from presto into velox: https://prestodb.io/docs/current/functions/aggregate.html#classification-metrics-aggregate-functions Classification functions all use `FixedDoubleHistogram`, which is a data structure to represent the bucket of weights. The index of the bucket for the histogram is evenly distributed between the min and value values. For all of the classification functions, the only difference is the extraction phase. All other steps will be the same. At a high level: - addRawInput will add a value into either the true or false weight bucket. The bucket to add the value to will depend on the prediction value. The prediction value is linearly mapped into a bucket based on (min, max and bucketCount) by normalizing the prediction between min and max. - The schema of the intermediate states is [version header][bucket count][min][max][weights] Differential Revision: D66684198
|
This pull request was exported from Phabricator. Differential Revision: D66684198 |
yuandagits
added a commit
to yuandagits/velox
that referenced
this pull request
Dec 11, 2024
Summary: Add the classification functions from presto into velox: https://prestodb.io/docs/current/functions/aggregate.html#classification-metrics-aggregate-functions Classification functions all use `FixedDoubleHistogram`, which is a data structure to represent the bucket of weights. The index of the bucket for the histogram is evenly distributed between the min and value values. For all of the classification functions, the only difference is the extraction phase. All other steps will be the same. At a high level: - addRawInput will add a value into either the true or false weight bucket. The bucket to add the value to will depend on the prediction value. The prediction value is linearly mapped into a bucket based on (min, max and bucketCount) by normalizing the prediction between min and max. - The schema of the intermediate states is [version header][bucket count][min][max][weights] Differential Revision: D66684198
yuandagits
force-pushed
the
export-D66684198
branch
from
6bf72c8 to
ef481a3
Compare
|
This pull request was exported from Phabricator. Differential Revision: D66684198 |
yuandagits
added a commit
to yuandagits/velox
that referenced
this pull request
Dec 13, 2024
Summary: Add the classification functions from presto into velox: https://prestodb.io/docs/current/functions/aggregate.html#classification-metrics-aggregate-functions Classification functions all use `FixedDoubleHistogram`, which is a data structure to represent the bucket of weights. The index of the bucket for the histogram is evenly distributed between the min and value values. For all of the classification functions, the only difference is the extraction phase. All other steps will be the same. At a high level: - addRawInput will add a value into either the true or false weight bucket. The bucket to add the value to will depend on the prediction value. The prediction value is linearly mapped into a bucket based on (min, max and bucketCount) by normalizing the prediction between min and max. - The schema of the intermediate states is [version header][bucket count][min][max][weights] Differential Revision: D66684198
yuandagits
force-pushed
the
export-D66684198
branch
from
ef481a3 to
1de9154
Compare
|
This pull request was exported from Phabricator. Differential Revision: D66684198 |
Yuhta
approved these changes
Dec 13, 2024
yuandagits
added a commit
to yuandagits/velox
that referenced
this pull request
Dec 14, 2024
Summary: Add the classification functions from presto into velox: https://prestodb.io/docs/current/functions/aggregate.html#classification-metrics-aggregate-functions Classification functions all use `FixedDoubleHistogram`, which is a data structure to represent the bucket of weights. The index of the bucket for the histogram is evenly distributed between the min and value values. For all of the classification functions, the only difference is the extraction phase. All other steps will be the same. At a high level: - addRawInput will add a value into either the true or false weight bucket. The bucket to add the value to will depend on the prediction value. The prediction value is linearly mapped into a bucket based on (min, max and bucketCount) by normalizing the prediction between min and max. - The schema of the intermediate states is [version header][bucket count][min][max][weights] Reviewed By: Yuhta Differential Revision: D66684198
yuandagits
force-pushed
the
export-D66684198
branch
from
1de9154 to
986bf72
Compare
|
This pull request was exported from Phabricator. Differential Revision: D66684198 |
Summary: Add the classification functions from presto into velox: https://prestodb.io/docs/current/functions/aggregate.html#classification-metrics-aggregate-functions Classification functions all use `FixedDoubleHistogram`, which is a data structure to represent the bucket of weights. The index of the bucket for the histogram is evenly distributed between the min and value values. For all of the classification functions, the only difference is the extraction phase. All other steps will be the same. At a high level: - addRawInput will add a value into either the true or false weight bucket. The bucket to add the value to will depend on the prediction value. The prediction value is linearly mapped into a bucket based on (min, max and bucketCount) by normalizing the prediction between min and max. - The schema of the intermediate states is [version header][bucket count][min][max][weights] Reviewed By: Yuhta Differential Revision: D66684198
yuandagits
force-pushed
the
export-D66684198
branch
from
986bf72 to
3c83e0e
Compare
|
This pull request was exported from Phabricator. Differential Revision: D66684198 |
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.
Summary:
Add the classification functions from presto into velox: https://prestodb.io/docs/current/functions/aggregate.html#classification-metrics-aggregate-functions
Classification functions all use
FixedDoubleHistogram, which is a data structure to represent the bucket of weights. The index of the bucket for the histogram is evenly distributed between the min and value values.For all of the classification functions, the only difference is the extraction phase. All other steps will be the same.
At a high level:
addRawInput will add a value into either the true or false weight bucket. The bucket to add the value to will depend on the prediction value. The prediction value is linearly mapped into a bucket based on (min, max and bucketCount) by normalizing the prediction between min and max.
The schema of the intermediate states is [version header][bucket count][min][max][weights]
Differential Revision: D66684198