crested.tl.losses.PoissonMultinomialLoss

crested.tl.losses.PoissonMultinomialLoss#

class crested.tl.losses.PoissonMultinomialLoss(total_weight=1.0, eps=1e-07, log_input=True, multinomial_axis='task', reduction='sum_over_batch_size', name='PoissonMultinomialLoss')#

Poisson decomposition with multinomial specificity term for aggregated counts.

Combines Poisson loss for total counts with a multinomial term for class proportions.

Parameters:

  • total_weight (float (default: 1.0)) – Weight of the Poisson term in the total loss.

  • eps (float (default: 1e-07)) – Small value to avoid log(0).

  • log_input (bool (default: True)) – If True, applies exponential transformation to predictions to produce counts.

  • multinomial_axis (str (default: 'task')) – Either “length” or “task”, representing the axis along which multinomial proportions are calculated.

  • reduction (str (default: 'sum_over_batch_size')) – Type of reduction to apply to the loss: “mean” or “none”.

  • name (str (default: 'PoissonMultinomialLoss')) – Name of the loss function.

Attributes table#

dtype

Methods table#

call(y_true, y_pred)

Compute the PoissonMultinomialLoss.

from_config(config)

get_config()

Return the configuration of the loss function.

Attributes#

PoissonMultinomialLoss.dtype#

Methods#

PoissonMultinomialLoss.call(y_true, y_pred)#

Compute the PoissonMultinomialLoss.

Parameters:

  • y_true – True target values (aggregated counts).

  • y_pred – Predicted values.

Returns:

Combined loss value.

classmethod PoissonMultinomialLoss.from_config(config)#
PoissonMultinomialLoss.get_config()#

Return the configuration of the loss function.