v0.9.0

Added

• Benchmark module (surfaces.benchmark) for structured optimizer comparison:
  • Benchmark class with budget management (Compute Units), multi-seed runs, and incremental execution
  • Trace storage and result aggregation via accessor pattern (results.summary, results.dataframe)
  • Statistical analysis: ranking, Critical Difference diagrams, multiple-comparison correction (Bonferroni, Holm, etc.)
  • Expected Running Time (ERT) computation
  • Progress bars and callback hooks during runs
  • Parallel execution backend
  • Optimizer resolution via duck typing (works with GFO, Optuna, scipy interfaces)
  • Error handling via catch-feature (failed evaluations do not abort the run)
• Multi-Fidelity support:
  • fidelity parameter on all ML test functions controls the fraction of training data used
  • _active_fidelity state for inspection during runs
  • Fidelity is forwarded through surrogate models for approximate low-fidelity evaluation
  • Hyperband / BOHB / ASHA compatible interface
• Compute Units (CU) as hardware-independent cost metric:
  • eval_cost attribute on every test function category
  • Calibration helper maps wall-clock times to CU
  • CU integration into CEC functions and into collection.filter
• Multi-objective test functions (new module algebraic.multi_objective):
  • ZDT1, ZDT2, ZDT3, ZDT4, ZDT6
  • DTLZ1 through DTLZ7
  • WFG1 through WFG9 with shared transformation primitives
  • n_objectives is now an instance parameter
• Discrete / combinatorial test functions (new module algebraic.discrete):
  • OneMaxFunction, LeadingOnesFunction, NKLandscapeFunction, TrapFunction, KnapsackFunction
  • Shared _base_discrete_function base class
• ML test function: CatBoostClassifierFunction (#19)
• Algebraic 4D test function: ColvilleFunction (closes #12, #18)
• Simulation robustness: FuturesTimeoutError handling for long-running ODE integrations
• Surrogate infrastructure:
  • _dependencies attribute on test functions so the surrogate collector can resolve required inputs
  • Dependency-aware training-data collection
  • Fidelity passthrough on pretrained ONNX surrogates
  • _get_training_data hook for custom trainers
  • Improved surrogate-trainer CLI (progress, collection scripts)
• Docs:
  • User guide for Benchmarking
  • User guide for Multi-Fidelity
  • User guide for Compute Units
  • Docstrings added across new modules

Changed

• Reworked benchmark API (pre-0.9 prototype is not compatible)
• Renamed metric parameter to objective across benchmark and statistics modules
• Moved custom test function module to its new location
• BBOB internals:
  • x_global attribute added and used consistently
  • _lambda_scale method replaces ad-hoc scaling
  • L attribute now drives rotation handling
  • RNG-based attribute initialization reworked for reproducibility
• CEC function calculation fixes across multiple functions
• Removed a duplicate optional-dependency entry in pyproject.toml
• Updated README with new examples and benchmark quickstart
• Updated CITATION.cff
• Replaced a deprecated upstream method in simulation code

Fixed

• Multiple correctness fixes in BBOB test functions
• Multiple correctness fixes in CEC functions
• RLCCircuitFunction numerical calculation
• Simulation test flakiness

What's Changed

• [Feature] Add Colville test function (closes #12) by @Jitenndra03 in #18
• Add CatBoost tabular ML test function by @3kyou1 in #19

New Contributors

• @Jitenndra03 made their first contribution in #18
• @3kyou1 made their first contribution in #19

Full Changelog: v0.8.1...v0.9.0

v0.8.1

Added

• Ensemble test functions: StackingEnsembleFunction, VotingEnsembleFunction, WeightedAveragingFunction
• BaseTabularEnsemble base class for ensemble methods
• MutualInfoFeatureSelectionFunction for feature selection optimization
• PolynomialFeatureTransformationFunction for polynomial feature engineering
• FeatureScalingPipelineFunction for scaling pipeline optimization
• XGBoostClassifierFunction for tabular ML
• LightGBMRegressorFunction and LightGBMClassifierFunction
• ShekelFunction (algebraic ND)
• Tests for CEC 2013 and CEC 2017 benchmarks
• Interface compliance and template method contract tests
• ML test function discovery for automated test coverage
• Developer guide in documentation
• CONTRIBUTING.md with contribution guidelines

Changed

• Refactored test functions into template method pattern
• Reworked test-function methods into accessor pattern
• Changed check_dependencies to decorator pattern
• Names are now generated automatically at class definition time (removed _name_ attribute)
• Reworked how optional dependencies are handled
• Moved default_bounds to spec
• Refactored multi-objective test functions
• Readded objective parameter to multi-objective functions

Fixed

• Fixed example in README
• Fixed documentation links

What's Changed

• Add LightGBM Regressor and Classifier test function by @ZohaibHassan16 in #14
• Add shekel funcfix by @ZohaibHassan16 in #16
• [Feature] Add XGBoost classifier test-function for tabular ML by @Shruti1128 in #17

New Contributors

• @ZohaibHassan16 made their first contribution in #14
• @Shruti1128 made their first contribution in #17

Full Changelog: v0.7.1...v0.8.1

v0.7.1

Fixed

• Windows-specific path handling error
• Bug in surrogate models
• Storage test issues

Changed

• Restructured ML test function directory layout
• Updated pre-commit configuration

Full Changelog: version/0.4.0...v0.7.1

v0.4.0

version/0.4.0

delete file