featran

A Scala feature transformation library for data science and machine learning

APACHE-2.0 License

Stars
466
Committers
27

Bot releases are hidden (Show)

featran -

Published by nevillelyh over 6 years ago

  • XGBoost LabeledPoint support
featran -

Published by nevillelyh over 6 years ago

  • Java TensorFlow Example extractor
featran -

Published by nevillelyh almost 7 years ago

  • Minimum Description Length (MDL) transformer
featran -

Published by nevillelyh almost 7 years ago

  • FeatureSpec composition
featran -

Published by nevillelyh almost 7 years ago

  • Java feature crossing
featran -

Published by nevillelyh almost 7 years ago

  • Performance improvements for streaming & backend use cases
  • Quantile and Inter Quartile Range Rejectors
featran -

Published by nevillelyh almost 7 years ago

  • Performance optimizations
  • Fix TensorFlow UnsupportedOperationException
featran -

Published by nevillelyh almost 7 years ago

  • Add n-grams transformer
  • Fix Java sparse array concurrency issue
  • Fix missing rejections in FeatureBuilder
  • Remove commons-math3 dependency
featran -

Published by nevillelyh about 7 years ago

  • Add feature crossing support
  • Add HeavyHitters transformer
  • Fix Java types in FloatSparseArray and DoubleSparseArray
featran -

Published by nevillelyh about 7 years ago

  • Scala 2.12 support (except Flink and Spark)
  • Add Java sparse array support
  • Update Algebird to 0.13.2
  • Use relatively safe max size of Java array
featran -

Published by nevillelyh about 7 years ago

  • Support feature rejection
  • Performance tweaks for MultiFeatureSpec
  • Java API
featran -

Published by nevillelyh about 7 years ago

  • fix NPE in MultiFeatureExtractor
  • minor performance tweaks
featran -

Published by fallonchen about 7 years ago

  • add hash one-hot, n-hot, n-hot-weighted encoders + performance improvements on *-hot encoders
  • support multiple specs and separate on output
featran -

Published by nevillelyh about 7 years ago

  • initial release
featran -

Published by nevillelyh about 7 years ago

  • add Flink support
  • add setting persistence
  • add normalizer
  • make bucketizer inclusive on upper bound
  • generalize float and double for feature builder
  • make Scalding and Spark dependencies provided
  • bug fixes and minor tweaks
featran -

Published by nevillelyh about 7 years ago

  • fix out of bound cases in min-max scaler
  • fix double overflow in quantile discretizer
  • add a method that keeps the original record after extraction
featran -

Published by nevillelyh about 7 years ago

  • add NumPy file support
  • add TensorFlow Example feature builder
  • support mapping features into previously stored settings
featran -

Published by nevillelyh about 7 years ago

  • add von Mises transformer
  • add n-hot weighted encoder
  • add vector identity transformer
  • add array collection type
  • bug fixes and minor tweaks
featran -

Published by nevillelyh about 7 years ago

  • 1-hot, n-hot and n-hot weighted encoder performance improvements