matchzoo.losses package

Submodules

matchzoo.losses.rank_cross_entropy_loss module

The rank cross entropy loss.

class matchzoo.losses.rank_cross_entropy_loss.RankCrossEntropyLoss(num_neg=1)

基类：object

Rank cross entropy loss.

Examples

>>> from keras import backend as K
>>> softmax = lambda x: np.exp(x)/np.sum(np.exp(x), axis=0)
>>> x_pred = K.variable(np.array([[1.0], [1.2], [0.8]]))
>>> x_true = K.variable(np.array([[1], [0], [0]]))
>>> expect = -np.log(softmax(np.array([[1.0], [1.2], [0.8]])))
>>> loss = K.eval(RankCrossEntropyLoss(num_neg=2)(x_true, x_pred))
>>> np.isclose(loss, expect[0]).all()
True

matchzoo.losses.rank_hinge_loss module

The rank hinge loss.

class matchzoo.losses.rank_hinge_loss.RankHingeLoss(num_neg=1, margin=1.0)

基类：object

Rank hinge loss.

Examples

>>> from keras import backend as K
>>> x_pred = K.variable(np.array([[1.0], [1.2], [0.8], [1.4]]))
>>> x_true = K.variable(np.array([[1], [0], [1], [0]]))
>>> expect = ((1.0 + 1.2 - 1.0) + (1.0 + 1.4 - 0.8)) / 2
>>> expect
1.4
>>> loss = K.eval(RankHingeLoss(num_neg=1, margin=1.0)(x_true, x_pred))
>>> np.isclose(loss, expect)
True

Module contents