def test_approx_ndcg_loss(self):
with tf.Graph().as_default():
scores = [[1.4, -2.8, -0.4], [0., 1.8, 10.2], [1., 1.2, -3.2]]
# ranks= [[1, 3, 2], [3, 2, 1], [2, 1, 3]]
labels = [[0., 2., 1.], [1., 0., 3.], [0., 0., 0.]]
weights = [[2.], [1.], [1.]]
example_weights = [[1., 2., 3.], [4., 5., 6.], [7., 8., 9.]]
norm_wts = [sum([wt * l for wt, l in zip(wts, lbls)]) / sum(lbls)
if sum(lbls) else 0
for wts, lbls in zip(example_weights, labels)]
with self.cached_session():
self.assertAlmostEqual(
ranking_losses._approx_ndcg_loss(labels, scores).eval(),
-((1 / (3 / ln(2) + 1 / ln(3))) * (3 / ln(4) + 1 / ln(3)) +
(1 / (7 / ln(2) + 1 / ln(3))) * (7 / ln(2) + 1 / ln(4))),
places=5)
self.assertAlmostEqual(
ranking_losses._approx_ndcg_loss(labels, scores, weights).eval(),
-(2 * (1 / (3 / ln(2) + 1 / ln(3))) * (3 / ln(4) + 1 / ln(3)) + 1 *
(1 / (7 / ln(2) + 1 / ln(3))) * (7 / ln(2) + 1 / ln(4))),
places=5)
self.assertAlmostEqual(
ranking_losses._approx_ndcg_loss(labels, scores,
example_weights).eval(),
-(norm_wts[0] * (1 / (3 / ln(2) + 1 / ln(3))) *
(3 / ln(4) + 1 / ln(3)) +
norm_wts[1] * (1 / (7 / ln(2) + 1 / ln(3))) *
(7 / ln(2) + 1 / ln(4))),
places=5)
def test_approx_ndcg_loss(self):
scores = [[1.4, -2.8, -0.4], [0., 1.8, 10.2], [1., 1.2, -3.2]]
labels = [[0., 2., 1.], [1., 0., 3.], [0., 0., 0.]]
weights = [[2.], [1.], [1.]]
with self.cached_session():
self.assertAlmostEqual(
ranking_losses._approx_ndcg_loss(labels, scores).eval(),
-((1 / (3 / ln(2) + 1 / ln(3))) * (3 / ln(4) + 1 / ln(3)) +
(1 / (7 / ln(2) + 1 / ln(3))) * (7 / ln(2) + 1 / ln(4))),
places=5)
self.assertAlmostEqual(
ranking_losses._approx_ndcg_loss(labels, scores, weights).eval(),
-(2 * (1 / (3 / ln(2) + 1 / ln(3))) * (3 / ln(4) + 1 / ln(3)) + 1 *
(1 / (7 / ln(2) + 1 / ln(3))) * (7 / ln(2) + 1 / ln(4))),
places=5)