29

• Add TH_PolyReg_IPOP_CMA_ES optimizer.
• Add Normalized MLP surrogate objective.
• Switch dependency manager to UV.
• Switch linters and formatters to ruff and ty.
• Make the CLI functionality a class.

28

• Use float64 everywhere for better precision.
• Optimized Bounds class to use numpy for some operations.
• Fixed numpy typehints related to numpy arrays.
• Add XGB surrogate objective function.
• Add MLP surrogate objective function.
• Silence numpy overflow and zero division warnings in places where it cannot be fixed.
• Add fallback in LWPR where the bandwidth is close to zero.
• Add top half metamodel and top half ipop cma es variation.
• Add lmm_ipop_cma_es.
• Use | None type annotation for all optional fields.
• Remove redundant type annotations from docstrings.

27

• Add the option to use zstandard compression in result pickles.

26

• KNN-CMA-ES and KNN-IPOP-CMA-ES now always make sure that the buffer size for KNN is at least the size of K.

25

• Optimize the approximate ranking metamodel to work an order of magnitude faster.
• Add remove_x methods to OptimizationRun, PointList and Point classes that substitute points' x values with None to save space.

24

• Add IPOP-CMA-ES and KNN-IPOP-CMA-ES.
• Add __len__ method to Bounds class.
• Refactor CmaEs helper functions.

23

• Count improvements and worsenings in aggregated pvalues table.
• Make optilab callable without python -m.
• Store version number in one place.

22

• Fix box plot layout so that it doesn't cut off the title.
• Add option to aggregate pvalues for all pickle files that are being processed.
• Add option to aggregate median and iqr for all pickle files that are being processed.
• Save stat test results to CSV files.
• Add argument to specify save path for plots and stats.
• Add argument to disable saving of any artifacts.

21

• Used faiss index in KNN surrogate function and locally weighted polynomial regression surrogate function for 10x performance boost.

20

• Move Mann-Whitney U test from __main__ to utils.stat_test.
• Separate Mann-Whitney U test function from tabulating function.
• Add second tutorial notebook about making custom objective functions and optimizers.
• Allow for plotting box plots without the outliers.
• Change ObjectiveFunction to store all metadata as member, remove get_metadata method.
• Overload slicing in PointList class.
• Display values below tolerance threshold as tolerance value threshold, allow to override this rule with --raw_values flag.
• Use deepcopy in bounds handling to avoid points being modified in place.
• Move common pylint disables to pyproject.toml.
• Add formatter for pyproject.toml.
• Add titles to plots containing type of plot and name of the function.
• Tilt x labels in box plots by 45 degrees so that they don't overlap.
• Add functionality to CLI tool to select entires with --entries argument.