def get_low_selective_features(csr_matrix, label, num_classes_feature_can_appear):
transposed = csr_matrix.transpose().tocsr()
results = parallel.rowwise(
transposed,
is_low_selective_,
min(8, transposed.shape[0]),
{"num_classes_feature_can_appear": num_classes_feature_can_appear, "label": label},
)
superset = set()
for r in results:
superset = superset.union(r)
return superset
def compute_row_sum_parallel(csr_matrix, parallelism):
results = parallel.rowwise(csr_matrix, row_sum_, parallelism, {})
superdict = dict()
for r in results:
superdict = dict(superdict.items() + r.items())
return superdict
