Lale

pretty-print bug fixes, added TopKVotingClassifier, added support for new libraries such as autoai_libs and imbalanced-learn, updated autopen schemas for documentation and more documentation enhancements.

Besides the capabilities from v0.3.4.6, this release makes it easier to use classifiers in the middle of a pipeline by supporting decision_function.

This release includes new features such as support for lists and tuples for nested higher order operators, schema validation changes to expect separate schemas for transform and predict and to allow more flexible schemas ('laleType':'any'), documentation updates and ability to generate
doc strings from schemas, fixes and updates to dependencies, fixes and updates to import and export from and to scikit-learn pipelines. It also includes bug fixes and enhancements to visualization, pretty_print, operator schemas, optimizer wrappers etc.

This release makes convenience methods (remove_last, visualize, pretty_print, transform_schema) available on all operators (individual, pipeline, choice). Also, this release makes improvements to higher-order operators (such as ensembles) and grammars.

This release turns search tools into proper Lale operators, so they work more smoothly with wrap_imported_operators and auto_configure. Also, this release expands the use of (sub-)schema checking on intermediate datasets.

Changes dependencies on liac-arff and mypy.