config.py

##################
# Settings File
##################

##
# 1. Experiment-specific properties
##
# 1.1. Comparison experiment
# 1.1.1. Bayesian optimization max iterations
n_calls = 200
# 1.1.2 Bayesian optimization convergence: stop optimization after n iterations without new optimum
n_convergence = 50

# 1.2. Iteration Experiment
# 1.2.1. Maximum number of iterations
max_calls = 200



##
# 2. General properties
##
# 2.1. Number of processes for parallelization
n_processes = 4

# 2.2. Number of splits for cross-validation (outside)
n_splits = 3

# 2.3. number of cross-validation splits inside of bayesian optimization
n_splits_bay_opt = 2

# 2.4. Number of features to be selected (default: only use 20 features)
min_nr_features = 20 # minimum number of features
max_nr_features = 20 # maximum number of features
iter_step_nr_features = 5 # step size

# 2.5 Number of points to evaluate of the acquisition function
n_acq_points = 5000

# 2.5. Datasets used
# numbers represent openml.org dataset ids
# booleans represent if a dataset is used for the experiment or not
data_ids = {
    "classification": {
        # experiment datasets
        12: False,
        312: False,
        316: False,
        851: False,
        978: False,
        1038: False,
        1039: False,
        1041: False,
        1042: False,
        1233: False,
        1468: False,
        1485: False,
        1501: False,
        1514: False,
        1515: False,
        4134: False,
        #40588: False, # more than 1 target
        #40592: False, # more than 1 target
        #40593: False, # more than 1 target
        #40594: False, # more than 1 target
        #40595: False, # more than 1 target
        #40596: False, # more than 1 target
        #40597: False, # more than 1 target
        40665: False,
        40910: False,
        40979: False,
        41083: False,
        41144: False,
        41158: False,
        41703: False,
        41939: False,
        41964: False,
        41966: False,
        41967: False,
        41973: False,
        # very small test dataset
        #1510: False # 30 features
    }
}

# 2.6 Datset column drop list (these columns are removed if they appear in a dataset)
drop_list = ["runstatus", "instance_id"]