This Fiddler object is used to define custom features based on the columns of a dataset. Custom features are mainly used for monitoring multi-dimensional vectors in NLP and CV use cases. This object holds information about which data columns constitute a custom feature, and information that specifies the settings for monitoring this custom feature.

AttributeTypeDefaultDescription
namestrNoneThe name of the custom feature
columns[srt]NoneA list of column name in the dataset which define this custom feature.
typeNone
transformationstrNone
n_clustersint10
monitorboolTrue

In order to initiate a custom feature whose elements are stored across columns of the baseline DataFrame, use the fdl.CustomFeature.from_columns() constructor function.