lir 1.9.0

v1.9.0

on Python PyPI

New Features

• Introduce new metrics (mean LLR, counts) by @PimMeulensteen in #478
• Leave two source out / 'population' to 'source_id' in SynthesizedNormalMulticlassData by @PimMeulensteen in #434
• Automatic schema generation by @wowtor in #480
• Aggregation group by @wowtor in #497
• add auc and equal_error_rate metrics by @wowtor in #513
• add count_sources metric for data statistics by @wowtor in #514
• Config parser for data setup can be specified by @wowtor in #501

Improvements

• Rename specific_source to binary_lrsystem and update references by @PimMeulensteen in #489
• Rename random_state to seed by @PimMeulensteen in #490
• Add a Automatic Speaker Recognition example by @PimMeulensteen in #492
• algorithms: replace SVM by calibrated classifier by @wowtor in #515
• Rename "data strategies" to the more specific "splitting strategies" by @wowtor in #518

Bug Fixes

• InstanceData: labels attribute should be used if hypothesis is None by @wowtor in #506
• parsing configuration for logging_pipeline raises index-out-of-range error if the context is the empty list by @wowtor in #511
• if ConfigValue.pop() raises an error, it should be of type YamlParseError by @wowtor in #517

Documentation

• New guide: extending lir by @wowtor in #436
• Rename section Registry reference to Built-in components by @wowtor in #482
• add page on YAML configuration structure by @wautron in #521

Minor changes

• Annotated function can be called directly by @wowtor in #475
• config: create an alias lir.config.ConfigAttribute by @wowtor in #502
• config: the method ConfigValue.pop() always returns a value of type ConfigValue, even if there is no such property by @wowtor in #500
• update logging pipeline to ignore non-feature data unless explicitly specified by @wowtor in #508

Full Changelog: v1.8.0...v1.9.0