def regression_loss(predicty, ylabel, weight_map=None):
"""
Distance regression loss
:param predicty: prediction results
:param ylabel: ground truth
:return: regression loss
"""
loss = l2_loss(predicty, ylabel, weight_map)
return loss
def test_l2_loss(self):
# expected loss: (0.04 + 4 + 1) /2 = 2.52
# (note - not the mean, just the sum)
with self.test_session():
predicted = tf.constant([1.2, 1, 2],
dtype=tf.float32,
name='predicted')
gold_standard = tf.constant([1, 3, 3],
dtype=tf.float32,
name='gold_standard')
self.assertAlmostEqual(
l2_loss(predicted, gold_standard).eval(), 2.52)
def test_l2_loss(self):
# expected loss: (0.04 + 4 + 1) /2 = 2.52
# (note - not the mean, just the sum)
with self.test_session():
predicted = tf.constant(
[1.2, 1, 2],
dtype=tf.float32, name='predicted')
gold_standard = tf.constant(
[1, 3, 3],
dtype=tf.float32, name='gold_standard')
self.assertAlmostEqual(
l2_loss(predicted, gold_standard).eval(), 2.52)