def test_topk_slicing(self):
join_top_k = slicer.process(self.all_features, self.complete_x, self.loss, len(self.complete_x), self.y_test, self.errors,
self.debug, self.alpha, self.k, self.w, self.loss_type, self.b_update)
union_top_k = union_slicer.process(self.all_features, self.complete_x, self.loss, len(self.complete_x), self.y_test, self.errors,
self.debug, self.alpha, self.k, self.w, self.loss_type, self.b_update)
self.assertEqual(join_top_k.min_score, union_top_k.min_score)
print("check1")
self.assertEqual(join_top_k.keys, union_top_k.keys)
print("check2")
self.assertEqual(len(join_top_k.slices), len(union_top_k.slices))
print("check3")
idx = -1
for sliced in join_top_k.slices:
idx += 1
self.assertEqual(sliced.score, union_top_k.slices[idx].score)
print("check4")
# alpha is size significance coefficient
# verbose option is for returning debug info while creating slices and printing it
# k is number of top-slices we want
# w is a weight of error function significance (1 - w) is a size significance propagated into optimization function
# enumerator <union>/<join> indicates an approach of next level slices combination process:
# in case of <join> in order to create new node of current level slicer
# combines only nodes of previous layer with each other
# <union> case implementation is based on DPSize algorithm
if enumerator == "join":
slicer.process(all_features,
complete_x,
f_l2,
x_size,
y_test,
errors,
debug=debug,
alpha=alpha,
k=k,
w=w,
loss_type=loss_type,
b_update=b_update)
elif enumerator == "union":
union_slicer.process(all_features,
complete_x,
f_l2,
x_size,
y_test,
errors,
debug=debug,
alpha=alpha,
k=k,
n_estimators=10, n_jobs=2, oob_score=False, random_state=0,
verbose=0, warm_start=False)
# alpha is size significance coefficient
# verbose option is for returning debug info while creating slices and printing it
# k is number of top-slices we want
# w is a weight of error function significance (1 - w) is a size significance propagated into optimization function
# loss_type = 0 (l2 in case of regression model
# loss_type = 1 (cross entropy in case of classification model)
preds = clf.predict(x_test)
predictions = []
counter = 0
mistakes = 0
for pred in preds:
predictions.append((counter, pred))
if y_test[counter] != pred:
mistakes = mistakes + 1
counter = counter + 1
lossF = mistakes / counter
# enumerator <union>/<join> indicates an approach of next level slices combination process:
# in case of <join> in order to create new node of current level slicer
# combines only nodes of previous layer with each other
# <union> case implementation is based on DPSize algorithm
if enumerator == "join":
slicer.process(all_features, complete_x, lossF, x_size, complete_y, predictions, debug=debug, alpha=alpha, k=k,
w=w, loss_type=loss_type, b_update=b_update)
elif enumerator == "union":
union_slicer.process(all_features, complete_x, lossF, x_size, complete_y, predictions, debug=debug, alpha=alpha,
k=k, w=w, loss_type=loss_type, b_update=b_update)