def beam_search(data, target_class, time_budget=conf.TIME_BUDGET, enable_i=True, top_k=conf.TOP_K,
beam_width=conf.BEAM_WIDTH,
iterations_limit=conf.ITERATIONS_NUMBER,
theta=conf.THETA,
quality_measure=conf.QUALITY_MEASURE,
diverse=True):
items = extract_items(data)
begin = datetime.datetime.utcnow()
time_budget = datetime.timedelta(seconds=time_budget)
# candidate_queue = items_to_sequences(items)
candidate_queue = [[]]
sorted_patterns = PrioritySet(top_k, theta=theta)
nb_iteration = 0
while datetime.datetime.utcnow() - begin < time_budget and nb_iteration < iterations_limit:
beam = PrioritySet()
while (len(candidate_queue) != 0) and nb_iteration < iterations_limit:
seed = candidate_queue.pop(0)
children = compute_children(seed, items, enable_i)
for child in children:
if nb_iteration >= iterations_limit:
break
quality = compute_quality(data, child, target_class, quality_measure=quality_measure)
# sorted_patterns.add_preserve_memory(child, quality, data)
sorted_patterns.add(child, quality)
beam.add(child, quality)
nb_iteration += 1
if diverse:
candidate_queue = [j for i, j in beam.get_top_k_non_redundant(data, beam_width)]
else:
candidate_queue = [j for i, j in beam.get_top_k(beam_width)]
# print("Number iterations beam search: {}".format(nb_iteration))
return sorted_patterns.get_top_k_non_redundant(data, top_k)
def test_over_top_k():
priority = PrioritySet()
priority.add(frozenset([1, 2]), 0.1)
assert len(priority.get_top_k(2)) == 1