tf.keras.losses.CategoricalGeneralizedCrossEntropy

Computes the Generalized Cross Entropy loss between y_true & y_pred.

Inherits From: Loss

tf.keras.losses.CategoricalGeneralizedCrossEntropy(
    q=0.5,
    reduction='sum_over_batch_size',
    name='categorical_generalized_cross_entropy',
    dtype=None
)

Generalized Cross Entropy (GCE) is a noise-robust loss function that provides better robustness against noisy labels than standard cross entropy. It generalizes both cross entropy and mean absolute error through the parameter q, where values closer to 1 make the loss more robust to noisy labels.

Formula:

loss = (1 - p**q) / q

where p is the predicted probability for the true class and q is the noise parameter.

q Float in range (0, 1). It is the noise parameter. Controls the behavior of the loss:

  • As q approaches 0: Behaves more like cross entropy
  • As q approaches 1: Behaves more like mean absolute error Defaults to 0.5
reduction Type of reduction to apply to the loss. In almost all cases this should be "sum_over_batch_size". Supported options are "sum", "sum_over_batch_size", "mean", "mean_with_sample_weight" or None. "sum" sums the loss, "sum_over_batch_size" and "mean" sum the loss and divide by the sample size, and "mean_with_sample_weight" sums the loss and divides by the sum of the sample weights. "none" and None perform no aggregation. Defaults to "sum_over_batch_size".
name Optional name for the loss instance.
dtype The dtype of the loss's computations. Defaults to None, which means using keras.backend.floatx(). keras.backend.floatx() is a "float32" unless set to different value (via keras.backend.set_floatx()). If a keras.DTypePolicy is provided, then the compute_dtype will be utilized.

Example:

y_true = np.array([0, 1, 0, 1])
y_pred = np.array([[0.7, 0.3], [0.2, 0.8], [0.6, 0.4], [0.4, 0.6]])
keras.losses.CategoricalGeneralizedCrossEntropy()(y_true, y_pred)

  • Zhang, Sabuncu, 2018 ("Generalized Cross Entropy Loss for Training Deep Neural Networks with Noisy Labels")

dtype

Methods

call

View source

call(
    y_true, y_pred
)

from_config

View source

@classmethod
from_config(
    config
)

get_config

View source

get_config()

__call__

View source

__call__(
    y_true, y_pred, sample_weight=None
)

Call self as a function.