Update Data Transforms #9371
Update Data Transforms #9371
Conversation
JRAlexander
requested review from
CESARDELATORRE,
chris-lauren,
justinormont,
rpetrusha,
shauheen and
shmoradims
| @@ -1,136 +1,185 @@ | |||
| --- | |||
| title: Data transforms in ML.NET | |||
| title: Machine learning data transforms - ML.NET | |||
There was a problem hiding this comment.
Are we just looking to reduce the name 'transforms' since we're moving to the estimators API?
There was a problem hiding this comment.
But are transforms estimators? That seems confusing.
| | Transform | Definition | | ||
| | --- | --- | | ||
| | <xref:Microsoft.ML.Legacy.Transforms.CombinerByContiguousGroupId> | Groups values of a scalar column into a vector based on a contiguous group ID. | | ||
| | <xref:Microsoft.ML.Transforms.GroupTransform> | Groups values of a scalar column into a vector based on a contiguous group ID. | |
There was a problem hiding this comment.
Might recommend:
Short desc:
"Groups multiple contiguous rows of data in to a single row by forming a vector type of the original input column"
Long desc:
"The Group transform groups the consecutive rows that share the specified group key (or keys). Both group keys and the aggregated values can be of arbitrary non-vector types. The resulting data will have all the group key columns preserved, and the aggregated columns will become variable-length vectors of the original types."
There was a problem hiding this comment.
Perhaps we should change the upsteam text where this file is produced from?
There was a problem hiding this comment.
Yes, that should be filed as an issue on the dotnet/machinelearning repo as the API ref is auto-generated from the code XMLcomments.
| | <xref:Microsoft.ML.Runtime.ImageAnalytics.ImageResizerTransform> | Takes one or more ImageType columns and resizes them to the provided height and width.| | ||
| | <xref:Microsoft.ML.Transforms.Text.LatentDirichletAllocationTransformer> | Implements LightLDA, a state-of-the-art implementation of Latent Dirichlet Allocation.| | ||
| | <xref:Microsoft.ML.Transforms.LoadTransform> | Loads specific transforms from the specified model file. Allows for 'cherry picking' transforms from a serialized chain, or to apply a pre-trained transform to a different (but still compatible) data view. | | ||
| | <xref:Microsoft.ML.Transforms.Text.NgramExtractingTransformer> | Produces a bag of counts of ngrams (sequences of consecutive values of length 1-n) in a given vector of keys. It does so by building a dictionary of ngrams and using the id in the dictionary as the index in the bag. | |
There was a problem hiding this comment.
Should be in #Text processing and featurization
| @@ -1,136 +1,185 @@ | |||
| --- | |||
| title: Data transforms in ML.NET | |||
There was a problem hiding this comment.
Also, I'd reorganize the doc into the name space..
Eg:
OneHotEncodingTransformer should go in the section for Microsoft.ML.Transforms.Categorical, "# Categorical".
Revised based on feedback. Co-Authored-By: JRAlexander <JRAlexander@users.noreply.github.com>
| * [Schema](#schema) | ||
| * [Text processing and featurization](#text-processing-and-featurization) | ||
| * [Miscellaneous](#miscellaneous) | ||
| # Machine learning data transforms - ML.NET |
There was a problem hiding this comment.
I find it a bit odd to have the - showing up in H1. I'd make the H1 different here and not use -
Maybe the original H1 was good and the non-branded terms are already added to the title metadata for SEO purposes.
| | <xref:Microsoft.ML.Legacy.Transforms.LabelColumnKeyBooleanConverter> | Transforms the label to either key or bool (if needed) to make it suitable for classification. | | ||
| | <xref:Microsoft.ML.Legacy.Transforms.LabelIndicator> | Label remapper used by OVA. | | ||
| | <xref:Microsoft.ML.Transforms.LabelConvertTransform> | Converts labels. | | ||
| | <xref:Microsoft.ML.Transforms.LabelIndicatorTransform> | Remaps multiclass labels to binary T,F labels, primarily for use with OVA.| |
There was a problem hiding this comment.
so people know what T,F labels are?
|
Thanks, @mairaw! |
Update Data Transforms for latest release.
Internal Review URL
Fixes #9476, #9552