tfa.losses.PinballLoss

View source on GitHub

Computes the pinball loss between y_true and y_pred.

loss = maximum(tau * (y_true - y_pred), (tau - 1) * (y_true - y_pred))

In the context of regression, this loss yields an estimator of the tau conditional quantile.

See: https://en.wikipedia.org/wiki/Quantile_regression

Usage:

pinball = tfa.losses.PinballLoss(tau=.1)
loss = pinball([0., 0., 1., 1.], [1., 1., 1., 0.])

# loss = max(0.1 * (y_true - y_pred), (0.1 - 1) * (y_true - y_pred))
#      = (0.9 + 0.9 + 0 + 0.1) / 4

print('Loss: ', loss.numpy())  # Loss: 0.475

Usage with the compile API:

model = tf.keras.Model(inputs, outputs)
model.compile('sgd', loss=tfa.losses.PinballLoss(tau=.1))

tau (Optional) Float in [0, 1] or a tensor taking values in [0, 1] and shape = [d0,..., dn]. It defines the slope of the pinball loss. In the context of quantile regression, the value of tau determines the conditional quantile level. When tau = 0.5, this amounts to l1 regression, an estimator of the conditional median (0.5 quantile).
reduction (Optional) Type of tf.keras.losses.Reduction to apply to loss. Default value is AUTO. AUTO indicates that the reduction option will be determined by the usage context. For almost all cases this defaults to SUM_OVER_BATCH_SIZE. When used with tf.distribute.Strategy, outside of built-in training loops such as tf.keras compile and fit, using AUTO or SUM_OVER_BATCH_SIZE will raise an error. Please see https://www.tensorflow.org/alpha/tutorials/distribute/training_loops for more details on this.
name Optional name for the op.

References:

fn The loss function to wrap, with signature fn(y_true, y_pred, **kwargs).
reduction (Optional) Type of tf.keras.losses.Reduction to apply to loss. Default value is AUTO. AUTO indicates that the reduction option will be determined by the usage context. For almost all cases this defaults to SUM_OVER_BATCH_SIZE. When used with tf.distribute.Strategy, outside of built-in training loops such as tf.keras compile and fit, using AUTO or SUM_OVER_BATCH_SIZE will raise an error. Please see this custom training tutorial for more details.
name (Optional) name for the loss.
**kwargs The keyword arguments that are passed on to fn.

Methods

from_config

Instantiates a Loss from its config (output of get_config()).

Args
config Output of get_config().

Returns
A Loss instance.

get_config

Returns the config dictionary for a Loss instance.

__call__

Invokes the Loss instance.

Args
y_true Ground truth values. shape = [batch_size, d0, .. dN], except sparse loss functions such as sparse categorical crossentropy where shape = [batch_size, d0, .. dN-1]
y_pred The predicted values. shape = [batch_size, d0, .. dN]
sample_weight Optional sample_weight acts as a coefficient for the loss. If a scalar is provided, then the loss is simply scaled by the given value. If sample_weight is a tensor of size [batch_size], then the total loss for each sample of the batch is rescaled by the corresponding element in the sample_weight vector. If the shape of sample_weight is [batch_size, d0, .. dN-1] (or can be broadcasted to this shape), then each loss element of y_pred is scaled by the corresponding value of sample_weight. (Note ondN-1: all loss functions reduce by 1 dimension, usually axis=-1.)

Returns
Weighted loss float Tensor. If reduction is NONE, this has shape [batch_size, d0, .. dN-1]; otherwise, it is scalar. (Note dN-1 because all loss functions reduce by 1 dimension, usually axis=-1.)

Raises
ValueError If the shape of sample_weight is invalid.