Note: version 0.1.2 due to an issue with creating the 0.1.1 release.
**Full Changelog**: https://github.com/AstraZeneca/rexmex/compare/v_00100...v_00102
v_00100
What's Changed
* Use registry pattern for ClassificationMetricSet by cthoyt in https://github.com/AstraZeneca/rexmex/pull/40
* Coverage refactor, added CoverageMetricSet and CoverageScoreCard by kajocina in https://github.com/AstraZeneca/rexmex/pull/42
* Annotate redundant functions by cthoyt in https://github.com/AstraZeneca/rexmex/pull/41
v_00015
What's Changed
* Update name in citation by cthoyt in https://github.com/AstraZeneca/rexmex/pull/39
* Add binarize annotation by cthoyt in https://github.com/AstraZeneca/rexmex/pull/36
* Cleanup scorecard interface by cthoyt in https://github.com/AstraZeneca/rexmex/pull/38
* Improve testing by cthoyt in https://github.com/AstraZeneca/rexmex/pull/37
**Full Changelog**: https://github.com/AstraZeneca/rexmex/compare/v_00014...v_00015
v_00014
What's Changed 🦖🦖
* Demonstrate annotating structured metadata to classification functions by cthoyt in https://github.com/AstraZeneca/rexmex/pull/29
* Add additional classification function annotations by cthoyt in https://github.com/AstraZeneca/rexmex/pull/35
**Full Changelog**: https://github.com/AstraZeneca/rexmex/compare/v_00013...v_00014
v_00013
- New method to filter training, testing, and validation instances.
- Exposing the get_performance_metrics() method.
v_00012
Fixes the import __all__
v_00011
New ranking metrics added are:
- Geometric Mean Rank
- Mean Rank
- Mean Reciprocal Rank
v_00010
Adds the following ranking metrics
- Reciprocal Rank
- Mean Reciprocal Rank (MRR)
- Average PrecisionK (APK)
- Mean Average PercisionK (MAPK)
- Average RecallK (ARK)
- Mean Average RecallK (MARK)
- HITSK
- Spearman's Rho
- Kendall Tau
- Intra List Similarity
- Personalization
- Novelty
- NPDM
- DCG
- NDCG
v_00009
- The metric sets on scorecards are exposed.
v_0007
- Fixes Broken Name Spacing
v_00008
- Improved aggregation behaviour.
v_00006
The new release separates metrics and creates namespaces based on metric categories. This helps with modularity and organization.
Results in namespaces for:
- Rating
- Classification
- Ranking
- Coverage
v_00005
Library now includes:
- Positive and negative likelihood ratio
- Informedness and markedness
- Threat score and critical success index
- Fowlkes - Mallows index
- Prevalence threshold
- Diagnostic odds ratio
v_00004
The new release covers:
- False Negative/Positive
- True Positive/Negative
- FPR, TPR, FNR, TNR
- Specificity, Selectivity, False Omission Rate, False Discovery Rate
- Miss Rate, Fall Out
- Positive Predictive Value, Negative Predictive Value
v_00003
New features and bug fixes:
- Normalization of targets
- Metric set behaviour changed
- New dataset for testing
- Completed test coverage
- Updated setup with tags and licensing
v_00001
This is a test release.