def run_on_real_dataset(dataset_dir):
"""`dataset_dir` should contain a file `*.data` in NumPy format."""
if os.path.isdir(dataset_dir):
data_files = [
x for x in os.listdir(dataset_dir) if x.endswith('.data')
]
if len(data_files) != 1:
raise ValueError("The dataset directory {} ".format(dataset_dir) +
"should contain one `.data` file but got " +
"{}.".format(data_files))
data_file = data_files[0]
data_path = os.path.join(dataset_dir, data_file)
name_expe = data_file.split('.')[0]
# Load real dataset and binarize
perfs = binarize(np.loadtxt(data_path))
# da_matrix = DAMatrix(perfs=perf, name=name_expe)
da_matrix = DAMatrix.load(dataset_dir)
da_matrix.perfs = perfs
meta_learners = get_the_meta_learners(
exclude_optimal=True)[1:] # Exclude random
run_expe(da_matrix,
meta_learners=meta_learners,
name_expe=name_expe,
with_once_random=True,
show_legend=True)
else:
raise ValueError("Not a directory: {}".format(dataset_dir))
def test_nfldamatrix():
da_matrix = NFLDAMatrix()
path_to_dir = da_matrix.save()
da_matrix2 = DAMatrix.load(path_to_dir)
print(da_matrix.perfs)
print(da_matrix2.perfs)