Skip to content

Latest commit

 

History

History
321 lines (108 loc) · 9.24 KB

CHANGELOG.md

File metadata and controls

321 lines (108 loc) · 9.24 KB

Changelog

All notable changes to this project will be documented in this file.

[unreleased]

Bug Fixes

  • Fix tests (#684)

  • Fixed missing runCounterfactuals in CFTest

  • Fixed bug attaching solverconfig to cfconfig

  • Fixed infinite look in cf generator (#33)

  • Shap Kernel Explainer (#874)
  • CF: TrustyService handling of CF resultsets (#799)
  • CF: ExplainabilityService handling of CF requests (#752)
  • MatrixUtils Extensions (#793)
  • CounterfactualExplainer: Add CF support for composite features (#1236)
  • Use constant values for constrained features in CF search (#822)
  • Counterfactual results: Limit number returned? (#888)
  • Duplicate counterfactual result id created for final result (#887)
  • Invalid CF results not being filtered out from consumers (#873)
  • Implement outcome match threshold for numerical features (#858)
  • Invalid size comparison between CF goals and model outputs (#884)
  • Implement Gower distance as the score metric for CFs (#911)
  • SHAP Error Bounds (#966)
  • Use configuration class instead of Builder for CFs (#967)
  • Fix flaky test TrafficViolationDmnLimeExplainerTest.testExplanationWeightedStabilityWithOptimization (#988)
  • Return ShapResults from SHAP that contains model null along with shap values (#989)
  • CF UI - Configuration of progress bar duration (#1020)
  • Increase CF test coverage by including DMN models (#1042)
  • Add support in CF for remaining ordinal feature types (#1229)
  • Support for text features in CF goals (#1034)
  • Handling of null values in CF score calculation (#1055)
  • SHAP batching options (#1264)
  • LARSPath and LassoLarsIC for future SHAP regularization (#1098)
  • Move SHAP to Apache Commons MatrixUtils (#1192)
  • Add initial support for remaining input features in CF as constrained (#1091)
  • Explainability: Native build is broken (#1143)
  • Convert WLR to use Apache MatrixUtils (#1181)
  • Integration of regularizers into SHAP (#1209)
  • Fix edge case in CF primary soft score calculation (#1208)
  • Scaled back the SHAP tests causing heap space issues #1240
  • Fix SHAP subset sampler out-of-range bug (#1249)
  • Add per-feature filtering to Datasets (#1266)
  • Add counterfactual entity factory case for Long feature values (#1270)
  • SHAPResults toString function (#1275)
  • Background generators first commit. CF generator is failing tests
  • Optimize test suite for faster builds
  • Added toString to all FeatureDomains and Explainers
  • Move arrow-converters into exp-core (#34)
  • Update Kogito and OptaPlanner versions (#38)
  • Swapped DummyModels to use more predictable models (#36)
  • Prepare for JBoss deployment (#40)
  • Add argument for Java 17 JPMS Arrow tests exception (#43)
  • Added kPerGoal flag to chained counterfactual generation (#47)
  • Add initial Dataframe implementation (#48)
  • Initial implementation of data-based fairness metrics (#51)
  • Connect Trusty service and Explainability service (#410)

Refactor

  • Refactored ShapResults to return dict of outputname, saliency like LIME

Security

  • Weighted random sampler (#755)

Testing

  • Tested

  • Tested shapresultstest

  • Tested after removing contamination