A comprehensive set of fairness metrics for datasets and machine learning models, explanations for these metrics, and algorithms to mitigate bias in datasets and models.
FACTS
/FACTS_bias_scan
Full Changelog: https://github.com/Trusted-AI/AIF360/compare/v0.6.0...v0.6.1
SenSeI
/SenSR
DeterministicReranking
ot_distance
bias_scan
from aif360.metrics
/aif360.sklearn.metrics
pytest-cov
by @aitorres in https://github.com/Trusted-AI/AIF360/pull/412
if_delegate_has_method
with available_if
by @hoffmansc in https://github.com/Trusted-AI/AIF360/pull/511
requests
dependency by @hoffmansc in https://github.com/Trusted-AI/AIF360/pull/519
fairadapt.R
in package by @hoffmansc in https://github.com/Trusted-AI/AIF360/pull/520
Full Changelog: https://github.com/Trusted-AI/AIF360/compare/v0.5.0...v0.6.0
class_imbalance
, kl_divergence
, conditional_demographic_disparity
intersection
and one_vs_rest
meta-metricsprot_attr
instead of an index labelpredict_proba
to RejectOptionClassifier
smoothed_edf
, df_bias_amplification
class_imbalance
, kl_divergence
, conditional_demographic_disparity
intersection
, one_vs_rest
bias_scan
in aif360.metrics
to be deprecated next releaseFull Changelog: https://github.com/Trusted-AI/AIF360/compare/v0.4.0...v0.5.0
This is a major release containing a number of new features, improvements, and bugfixes.
compat.v1
) support added (#230)MetaFairClassifier
code cleaned and sped up (#196)maxiter
and maxfun
arguments in LFR fit()
(#184)scores
in a single-row dataset was getting squeezed (#193)consistency_score
documentation (#195)@baba-mpe, @SSaishruthi, @leenamurgai, @synapticarbors, @sohiniu, @yangky11
This is a major release containing a number of new features, improvements, and bugfixes.
pip install 'aif360[LFR,AdversarialDebiasing]'
or pip install 'aif360[all]'
scores
output to AdversarialDebiasing.predict()
(#139)subset()
method to StructuredDataset
(#140)MulticlassLabelDataset
to support basic multiclass problems (#165)statistical_parity_difference
, disparate_impact_ratio
, equal_opportunity_difference
, average_odds_difference
, average_odds_error
, between_group_generalized_entropy_error
)generalized_entropy_index
and its variants, consistency_score
)specificity_score
, base_rate
, selection_rate
, generalized_fpr
, generalized_fnr
)make_scorer
function to wrap metrics for use in sklearn cross-validation functions (#174, #178)Reweighing
, AdversarialDebiasing
, CalibratedEqualizedOdds
)NotImplementedError
in StandardDataset
(#115)GermanDataset
(#129 and #137)@autoih, @romeokienzler, @jimbudarz, @stephanNorsten, @sethneel, @imolloy, @guillemarsan, @gdequeiroz, @chajath, @bhavyaghai, @Tomcli, @swapna-somineni, @chkoar, @motapaolla
This is a major release containing a number of new features, improvements, and bugfixes.
pip install 'aif360[LFR,AdversarialDebiasing]'
or pip install 'aif360[all]'
scores
output to AdversarialDebiasing.predict()
(#139)subset()
method to StructuredDataset
(#140)statistical_parity_difference
, disparate_impact_ratio
, equal_opportunity_difference
, average_odds_difference
, average_odds_error
, between_group_generalized_entropy_error
)generalized_entropy_index
and its variants, consistency_score
)specificity_score
, base_rate
, selection_rate
, generalized_fpr
, generalized_fnr
)Reweighing
, AdversarialDebiasing
, CalibratedEqualizedOdds
)NotImplementedError
in StandardDataset
(#115)GermanDataset
(#129 and #137)@autoih, @romeokienzler, @jimbudarz, @stephanNorsten, @sethneel, @imolloy, @guillemarsan, @gdequeiroz, @chajath, @bhavyaghai, @Tomcli
New Algorithm:
DisparateImpactRemover
OptimPreproc
to use the latest version of cvxpy
labels
from predicted scores
in CalibratedEqOddsPostprocessing
scores_names
arg in StructuredDataset
allows for easier importing of predictions run elsewheretutorial_gender_classification
notebook now uses skimage
instead of cv2
aif360.__version__
now returns the correct version stringReweighing
; added new demo using AdversarialDebiasing
on Adult Datasetsubprocess.run
in PrejudiceRemover
for Python 2.7 compatibilitycategorical_features
would not take into account features_to_drop
in StandardDataset
@ckadner, @cclauss, @vijaykeswani, @ffosilva, @kant, @adrinjalali, @mariaborbones