def test_if_selected_close_to_other_both_umarked_tav_considered2(
self, cfg):
# given
cl = Classifier(quality=0.8, tav=0.1, cfg=cfg)
cl_del = Classifier(quality=0.85, tav=0.1, cfg=cfg)
# when
selected_cl = ClassifiersList(cfg=cfg)\
.select_preferred_to_delete(cl, cl_del)
# then
assert cl.is_marked() is False
assert cl_del.is_marked() is False
assert cl_del == selected_cl
def test_if_selected_somewhat_close_to_other_marked_considered2(self, cfg):
# given
cl = Classifier(quality=0.8, cfg=cfg)
cl_del = Classifier(quality=0.85, cfg=cfg)
cl_del.mark[0].add('0')
# when
selected_cl = ClassifiersList(cfg=cfg)\
.select_preferred_to_delete(cl, cl_del)
# then
assert cl.is_marked() is False
assert cl_del.is_marked() is True
assert cl_del == selected_cl
def expected_case(cl: Classifier, p0: Perception,
time: int) -> Optional[Classifier]:
"""
Controls the expected case of a classifier. If the classifier
is too specific it tries to add some randomness to it by
generalizing some attributes.
:param cl:
:param p0:
:param time:
:return: new classifier or None
"""
if cl.cfg.do_pee:
cl.effect.update_enhanced_effect_probs(p0, cl.cfg.beta)
diff = cl.mark.get_differences(p0)
if diff.specificity == 0:
if cl.cfg.do_pee and cl.is_marked():
cl.ee = True
cl.increase_quality()
return None
no_spec = len(cl.specified_unchanging_attributes)
no_spec_new = diff.specificity
child = cl.copy_from(cl, time)
if no_spec >= cl.cfg.u_max:
while no_spec >= cl.cfg.u_max:
res = cl.generalize_unchanging_condition_attribute()
assert res is True
no_spec -= 1
while no_spec + no_spec_new > cl.cfg.u_max:
if random() < 0.5:
diff.generalize_specific_attribute_randomly()
no_spec_new -= 1
else:
if cl.generalize_unchanging_condition_attribute():
no_spec -= 1
else:
while no_spec + no_spec_new > cl.cfg.u_max:
diff.generalize_specific_attribute_randomly()
no_spec_new -= 1
child.condition.specialize_with_condition(diff)
if child.q < 0.5:
child.q = 0.5
return child
