def test_should_return_abs_diff_for_single_value(self):
with tf.Graph().as_default():
labels = tf.constant([0.9])
outputs = tf.constant([0.1])
loss = l1_loss(labels, outputs)
with tf.Session() as session:
assert_close(session.run([loss])[0], 0.8)
def test_should_return_zero_if_logits_are_matching_labels_with_neg_pos_value(
self):
with tf.Graph().as_default():
labels = tf.constant([[[0.0, 1.0]]])
logits = tf.constant([[[-10.0, 10.0]]])
loss = cross_entropy_loss(labels, logits)
with tf.Session() as session:
assert_close(session.run([loss])[0], 0.0)
def test_evaluate_predictions():
n_classes = 4
predictions = tf.constant(np.array([0, 1, 1, 2, 3, 3]))
labels = tf.constant(np.array([0, 1, 2, 3, 3, 3]))
evaluation_tensors = evaluate_predictions(labels=labels,
predictions=predictions,
n_classes=n_classes)
with tf.Session() as session:
assert np.array_equal(
session.run(evaluation_tensors.confusion_matrix),
np.array([[1, 0, 0, 0], [0, 1, 0, 0], [0, 1, 0, 0], [0, 0, 1, 2]]))
assert np.array_equal(session.run(evaluation_tensors.tp),
np.array([1, 1, 0, 2]))
assert np.array_equal(session.run(evaluation_tensors.fp),
np.array([0, 1, 1, 0]))
assert np.array_equal(session.run(evaluation_tensors.fn),
np.array([0, 0, 1, 1]))
expected_micro_precision = 4.0 / (4 + 2)
expected_micro_recall = 4.0 / (4 + 2)
expected_micro_f1 = (
2 * expected_micro_precision * expected_micro_recall /
(expected_micro_precision + expected_micro_recall))
assert_close(session.run(evaluation_tensors.micro_precision),
expected_micro_precision)
assert_close(session.run(evaluation_tensors.micro_recall),
expected_micro_recall)
assert_close(session.run(evaluation_tensors.micro_f1),
expected_micro_f1)
def test_should_return_zero_value_for_not_occuring_class(self):
frequencies = [[1, 1], [1, 1], [0, 0]]
result = calculate_median_weights_for_frequencies(frequencies)
get_logger().debug('result: %s', result)
assert_close(sum(result), 1.0)
assert_all_close(result, [0.5, 0.5, 0.0], atol=0.001)