def find_set_covering(
subsets: List[CommonAncestor],
ontology: Ontology = None,
value: List[float] = None,
max_num_subsets: int = None
) -> Union[None, List[Tuple[str, Set[str]]]]:
"""greedy algorithm to solve set covering problem on subsets of trimming candidates
Args:
ontology: optional ontology to avoid parent-child relationships in the final result
subsets (List[Tuple[str, str, Set[str]]]): list of subsets, each of which must contain a tuple with the first
element being the ID of the subset, the second being the name, and the third the actual set of elements
value (List[float]): list of costs of the subsets
max_num_subsets (int): maximum number of subsets in the final list
Returns:
Union[None, List[str]]: the list of IDs of the subsets that maximize coverage with respect to the elements
in the element universe
"""
logger.debug("starting set covering optimization")
elem_to_process = {subset.node_id for subset in subsets}
if value and len(value) != len(elem_to_process):
return None
universe = set(
[e for subset in subsets for e in subset.covered_starting_nodes])
included_elmts = set()
included_sets = []
while len(elem_to_process) > 0 and included_elmts != universe and (
not max_num_subsets or len(included_sets) < max_num_subsets):
if value:
effect_sets = sorted([
(v * len(s.covered_starting_nodes - included_elmts),
s.covered_starting_nodes, s.node_label, s.node_id)
for s, v in zip(subsets, value) if s.node_id in elem_to_process
],
key=lambda x: (-x[0], x[2]))
else:
effect_sets = sorted(
[(len(s.covered_starting_nodes - included_elmts),
s.covered_starting_nodes, s.node_label, s.node_id)
for s in subsets if s.node_id in elem_to_process],
key=lambda x: (-x[0], x[2]))
elem_to_process.remove(effect_sets[0][3])
if ontology:
for elem in included_sets:
if effect_sets[0][3] in ontology.ancestors(elem[0]):
included_sets.remove(elem)
included_elmts |= effect_sets[0][1]
included_sets.append((effect_sets[0][3], effect_sets[0][1]))
logger.debug("finished set covering optimization")
return included_sets
def split_assocs(self, root_target : TargetClass, ontology : Ontology = None):
logging.info('Splitting assocs on: {} // {}'.format(root_target, ontology))
aset = self.assocs
if ontology is None:
ontology = aset.ontology
fmap = {}
tmap = {}
for subj in aset.subjects:
targets = set()
features = set()
for c in aset.annotations(subj):
if root_target in ontology.ancestors(c, reflexive=True):
targets.add(c)
else:
features.add(c)
fmap[subj] = features
tmap[subj] = targets
self.assocs = AssociationSet(ontology=ontology, association_map=fmap)
self.target_assocs = AssociationSet(ontology=ontology, association_map=tmap)
logging.info('Split; f={} t={}'.format(self.assocs, self.target_assocs))