def RFReprodExchTrees(list_indiv, max_id, options):
list_indiv = list(list_indiv)
clfp1 = list_indiv[0].clf
clfp2 = list_indiv[1].clf
clfchild = copy.deepcopy(clfp1)
if options['cross_shuffle']:
n1 = clfp1.n_estimators // 2
n2 = clfp2.n_estimators // 2
clf1_p = np.random.choice(clfp1.estimators_, size=n1, replace=False)
clf2_p = np.random.choice(clfp2.estimators_, size=n2, replace=False)
clfchild.estimators_ = list(clf1_p) + list(clf2_p)
else:
inds = np.random.randint(0, clfp2.n_estimators,
clfp2.n_estimators // 2)
for j in inds:
clfchild.estimators_[j] = clfp2.estimators_[j]
clfchild.n_estimators = len(clfchild.estimators_)
new_id = max_id + 1
indiv3 = genetic.individual(clfchild,
new_id,
type_rf=True,
alpha=options['alpha'],
evaluate_on_data=options['on_data'],
X=options['X'],
y=options['y'])
return indiv3
def RFReprodCrossOver(list_indiv, max_id, options):
list_indiv = list(list_indiv)
if len(list_indiv) > 2:
print('Warning : more than 2 parents..')
elif len(list_indiv) < 2:
print('Warning : parent missing..')
else:
clfp1 = list_indiv[0].clf
clfp2 = list_indiv[1].clf
clfchild = copy.deepcopy(clfp1)
new_id = max_id + 1
for k, e in enumerate(clfp1.estimators_):
###
node = np.random.randint(e.tree_.node_count)
###
clfchild.estimators_[k] = CoherentFusionDecisionTree(
e, node, clfp2.estimators_[k])
indiv3 = genetic.individual(clfchild,
new_id,
type_rf=True,
alpha=options['alpha'],
evaluate_on_data=options['on_data'],
X=options['X'],
y=options['y'])
return indiv3
def ReprodIndividualsFromRF(list_indiv, max_id, options):
list_indiv = list(list_indiv)
rf = RandomForestClassifier(n_estimators=len(list_indiv))
trees = list()
for indiv in list_indiv:
trees.append(indiv.clf)
rf.estimators_ = trees
rf.n_classes_ = trees[0].n_classes_
rf.classes_ = trees[0].classes_
new_dt = eqtree_rec_rf(rf,
0,
max_depth=options['max_depth'],
smallest_tree=False)
new_id = max_id + 1
indiv3 = genetic.individual(new_dt,
new_id,
type_rf=False,
alpha=options['alpha'],
evaluate_on_data=options['on_data'],
X=options['X'],
y=options['y'])
return indiv3
def ReprodCrossOver(list_indiv, max_id, options):
list_indiv = list(list_indiv)
if len(list_indiv) > 2:
print('Warning : more than 2 parents..')
elif len(list_indiv) < 2:
print('Warning : parent missing..')
else:
new_id = max_id + 1
###
node = np.random.randint(list_indiv[0].clf.tree_.node_count)
###
new_dt = CoherentFusionDecisionTree(list_indiv[0].clf, node,
list_indiv[1].clf)
indiv3 = genetic.individual(new_dt,
new_id,
type_rf=False,
alpha=options['alpha'],
evaluate_on_data=options['on_data'],
X=options['X'],
y=options['y'])
return indiv3
1,
L_c,
0.005,
EV_type='depth',
type_rf=True,
flow_weighted_costs=True,
on_data=True,
X=X_train_budget,
y=y_train_budget)
for i in range(N_rep):
ind_init = genetic.individual(clf_init,
0,
alpha,
type_rf=True,
evaluate_on_data=True,
X=X_train_budget,
y=y_train_budget)
init_value = ind_init.compute_value(budget)
init_depth = ind_init.compute_depth()
init_model_size = ind_init.compute_model_size()
pop = genetic.population(alpha,
type_rf=True,
evaluate_on_data=True,
X_source=X_train_init,
y_source=y_train_init,
X=X_train_budget,
y=y_train_budget,
budget_function=budget,
def test_individual():
"""Test individual generation"""
config = genetic.individual()
assert(isinstance(config, list))
assert(len(config) == genetic.genes)