Experiment Data for A Per-Class Automated Algorithm Selection Framework for General Black-Box Optimisation

Raw experimental output, and data organised for analysis code (to be published separately) in paths ending with "_organised". All in IOHprofiler format. Experimental settings: Algorithms: All NGOpt choices for nevergrad version 0.5.0 (total 39 algorithms; nevergrad version 0.6.0 uses a subset of 34 of these algorithms for the same settings): CMA, ChainMetaModelPowell, Cobyla, MetaModel, MetaModelOnePlusOne, ParametrizedMetaModel(multivariate optimizer=CmaFmin2), 33 different ConfPortfolio versions Data file specific content summary BBOB TRAIN data_seeds2.zip Raw results 24 BBOB problems, 25 runs (each with a different seed) on instance 1 for every problem Evaluation budget: 10000 Dimensions: 2, 3, 5, 10, 15, 20, 25, 30, 35, 40, 45, 50, 60, 70, 80, 90, 100 data_seeds2_organised.zip Data from data_seeds2.zip organised for use with analysis code data_seeds2_bud_dep.zip Raw results F1 (sphere) BBOB problem, 25 runs (each with a different seed) on instance 1 Evaluation budget: 200 Dimensions: 100 data_seeds2_bud_dep_organised.zip Data from data_seeds2_bud_dep.zip organised for use with analysis code data_seeds2_ngopt.zip Raw results Algorithms: NGOpt39's choice per budget-dimension combination (on Nevergrad version 0.6.0) 24 BBOB problems, 25 runs (each with a different seed) on instance 1 for every problem Evaluation budgets: 200, 300, 500, 1000, 1500, 2000, 2500, 3000, 3500, 4000, 4500, 5000, 6000, 7000, 8000, 9000, 10000 Dimensions: 2, 3, 5, 10, 15, 20, 25, 30, 35, 40, 45, 50, 60, 70, 80, 90, 100 data_seeds2_ngopt_organised.zip Data from data_seeds2_ngopt.zip organised for use with analysis code BBOB TEST data_seeds2_bbob_test.zip Raw results WARNING: Also contains preprocessed data from preprocessing during execution prior to a bugfix in the preprocessing code (raw data is valid and usable after reprocessing) Algorithms: NGOpt39's choice, and the remaining top four algorithms per budget-dimension combination (i.e., five algorithms per budget-dimension combination) 24 BBOB problems, 1 run on per instance in 2-26 for every problem Evaluation budgets: 200, 300, 500, 1000, 1500, 2000, 2500, 3000, 3500, 4000, 4500, 5000, 6000, 7000, 8000, 9000, 10000 Dimensions: 2, 3, 5, 10, 15, 20, 25, 30, 35, 40, 45, 50, 60, 70, 80, 90, 100 data_seeds2_bbob_test_reprocessed.zip Data from data_seeds2_bbob_test.zip organised and preprocessed data for use with analysis code; reprocessed after bugfix MA-BBOB (TEST) data_seeds2_ma_complete.zip Raw results WARNING: Also contains preprocessed data from preprocessing during execution prior to a bugfix in the preprocessing code (raw data is valid and usable after reprocessing) Algorithms: NGOpt39's choice, and the remaining top four algorithms per budget-dimension combination (i.e., five algorithms per budget-dimension combination) 828 MA-BBOB problems, 1 run each (all composed combinations of instance 1 from the 24 BBOB problems) Evaluation budgets: 200, 300, 500, 1000, 1500, 2000, 2500, 3000, 3500, 4000, 4500, 5000, 6000, 7000, 8000, 9000, 10000 Dimensions: 2, 3, 5, 10, 15, 20, 25, 30, 35, 40, 45, 50, 60, 70, 80, 90, 100 data_seeds2_ma_complete_reprocessed.zip Data from data_seeds2_ma_complete.zip organised and preprocessed data for use with analysis code; reprocessed after bugfix Package versions: Performance data for all problem instances and algorithms was collected with nevergrad version 0.6.0.