Permalink
Fetching contributors…
Cannot retrieve contributors at this time
82 lines (63 sloc) 3.56 KB

scikit-multilearn Changelog

0.1.0 (release planned 2018-06-14)

Fix a lot of bugs and generally improve stability, cross-platform functionality standard and unit test coverage. This release has been tested with a large set of unit tests that work across Windows

Also, new features:

  • multi-label stratification algorithm and stratification quality measures
  • a robust reorganization of label space division, alongside with a working stochastic blockmodel approach and new underlying layer - graph builders that allow using graph models for dividing the label space based not just on label co-occurence but on any kind of network relationships between labels you can come up with
  • meka wrapper works fully cross-platform now, including windows 10
  • multi-label data set downloading and load/save functionality brought in, like sklearn's dataset
  • kNN models support sparse input
  • neuro fuzzy models support sparse input
  • BSD-compatible label space partitioning via NetworkX
  • dependence on GPL libraries made optional
  • working predict_proba added for label space partitioning methods

0.0.5 (released 2017-02-25)

  • a general matrix-based label space clusterer has been added which can cluster the output space using any scikit-learn compatible clusterer (incl. k-means)
  • support for more single-class and multi-class classifiers you can now use problem transformation approaches with your favourite neural networks/deep learning libraries: theano, tensorflow, keras, scikit-neuralnetworks
  • support for label powerset based stratified kfold added
  • graph-tool clusterer supports weighted graphs again and includes stochastic blockmodel calibration
  • bugs were fixed in: classifier chains and hierarchical neuro fuzzy clasifiers

0.0.4 (released 2017-02-04)

  • *kNN classifiers support sparse matrices properly
  • support for the new model_selection API from scikit-learn
  • extended graph-based label space clusteres to allow taking probability of a label occuring alone into consideration
  • compatible with newest graphtool
  • support the case when meka decides that an observation doesn't have any labels assigned
  • HARAM classifier provided by Fernando Benitez from University of Konstanz
  • predict_proba added to problem transformation classifiers
  • ported to python 3

0.0.3 (released 2016-06-03)

  • support for new multi-label classification methods:

    • classsifier chains (CC)
    • multi-label kNN methods: BRkNN and MLkNN
    • all classifiers use sparse matrices internally
    • a general network for clustering label space with a flat classifier
    • the classifiers work with scikit pipelines / CVs
  • interface to meka 1.9, meka can work as a scikit-ml classifier

  • loading arff files to sparse matrices by default

0.0.2 (removed due to bug in PyPi)

  • support for new multi-label classification methods:
    • classsifier chains (CC)
    • multi-label kNN methods: BRkNN and MLkNN
    • all classifiers use sparse matrices internally
    • a general network for clustering label space with a flat classifier
    • the classifiers work with scikit pipelines / CVs

0.0.1 (released 2014-12-01)

  • initial release
  • support for initial set of multi-label classification methods:
    • binary relevance, label powerset
    • RAkEL both distinct and overlapping
    • label cooccurence based distinct partitioning classifiers
  • interface to meka 1.7
  • ARFF to numpy.array convertion classes and data set manipulation