Releases: sdv-dev/SDMetrics
Releases · sdv-dev/SDMetrics
v0.23.0
v0.22.0
New Features
- Add a fairness metric that computes Equalized Odds - Issue #772 by @fealho
- Add a violin plot visualizations to compare a pair of columns - Issue #759 by @R-Palazzo
Internal
- Fix test failures with pandas 2.3 - Issue #773 by @rwedge
- Check pyproject for release candidate dependencies - Issue #770 by @rwedge
Maintenance
- Update python setup step in workflows to use latest python version - Issue #653 by @frances-h
Miscellaneous
v0.21.0 - 2025-05-29
New Features
- Add a violin plot visualizations to compare a pair of columns - Issue #759 by @R-Palazzo
v0.20.1 - 2025-04-14
Bugs Fixed
- Importing sdmetrics crashes because of torch dependency - Issue #764 by @amontanez24
v0.20.0 - 2025-04-11
New Features
- Consolidate handling of datetime columns - Issue #741 by @frances-h
- Improve handling of datetime columns for metrics - Issue #740 by @lajohn4747
- DCR metrics should allow the possibility of constant numerical columns - Issue #739 by @lajohn4747
- Add
DCRBaselineProtection
metric - Issue #720 - Add
DCROverfittingProtection
metric - Issue #719 by @lajohn4747
Bugs Fixed
- Visualization
get_column_plot
crashes if I have a column of all null values - Issue #757 by @npatki - The
DCRBaselineProtection
metric is not creating the correct size of random data - Issue #743 by @lajohn4747 - The
DCRBaselineProtection
metric is not producing the expected score - Issue #742 by @lajohn4747 - The DCRBaselineProtection metric crashes when the distance between random data and real data is 0 - Issue #738 by @npatki
- DCROverfitting and DCRBaseline metrics produce too many warnings about missing columns. - Issue #737 by @lajohn4747
- Wrong default values for the
classifier
parameter of theBinaryClassifierPrecisionEfficacy
- Issue #734 by @R-Palazzo
Maintenance
- Upgrade
pomegranate
dependency to Python 3.13 - Issue #717 by @R-Palazzo
v0.19.0 - 2025-02-24
New Features
- Use metadata to validate inputs to
BinaryClassifierEfficacy
precision and recall metrics - Issue #727 by @R-Palazzo - Speed up calculation of the QualityReport - Issue #718 by @R-Palazzo
- Allow subsampling when computing the
ContingencySimilarity
metric - Issue #716 by @R-Palazzo - Add
BinaryClassifierRecallEfficacy
metric - Issue #712 by @R-Palazzo - Add
BinaryClassifierPrecisionEfficacy
metric - Issue #711 by @R-Palazzo - The
ContingencySimilarity
metric should be able to discretize continuous columns - Issue #700
Maintenance
- Delete conda folder - Issue #708 by @amontanez24
- Cancel previous workflow runs on a new push - Issue #705 by @R-Palazzo
- Support Python 3.13 (excluding
pomegranate
) - Issue #645 by @rwedge
Miscellaneous
- BNLikelihood - Issue #630
v0.18.0 - 2024-12-13
Bugs Fixed
- Missing whitespace in
DisclosureProtection
warning - Issue #694 by @frances-h DisclosureProtection
should be NaN if baseline score is zero - Issue #693 by @frances-hCategoricalCAP
metric returns 0 if no overlap in known fields - Issue #692 by @frances-h
New Features
- Add
DisclosureProtectionEstimate
metric - Issue #676 by @frances-h - Add
DisclosureProtection
metric - Issue #675 by @frances-h
v0.17.1 - 2024-12-04
Maintenance
- Create Prepare Release workflow - Issue #674 by @amontanez24
- Update codecov and add flag for integration tests - Issue #644 by @pvk-developer
Bugs Fixed
InterRowMSAS
ignores sequences with missing values - Issue #679 by @fealho- Improve error handling for datetime values when
apply_log = True
forInterRowMSAS
- Issue #672 by @fealho - Improve warning handling for non-positive values when
apply_log = True
forInterRowMSAS
- Issue #670 by @fealho StatisticMSAS
raises undesirableFutureWarning
- Issue #665 by @fealhoKSComplement
can be unstable for constant float values - Issue #652 by @fealho
v0.17.0 - 2024-11-14
This release adds a number of Multi-Sequence Aggregate Similarity (MSAS) metrics!
Bugs Fixed
- Relocate timeseries metrics modules - Issue #661 by @fealho
- Fix
SequenceLengthSimilarity
docstrings - Issue #660 by @fealho - When running Quality Report, ContingencySimilarity produces a RuntimeWarning (
The values in the array are unorderable.
) - Issue #656 by @R-Palazzo
New Features
v0.16.0 - 2024-09-25
This release improves the performance of the contingency_similarity
metric. It also factors dtypes into the score of the TableStructure
metric.
Internal
- Try to improve performance of contingency_similarity - Issue #622 by @amontanez24
New Features
- Add dtype comparison in
TableStructure
metric (used in Diagnostic report) - Issue #631 by @R-Palazzo