brainpy.losses.cross_entropy_sigmoid

brainpy.losses.cross_entropy_sigmoid(predicts, targets)

Computes the sigmoid cross-entropy loss.

Parameters

• predicts – (batch, …, #class) tensor of logits.

• targets – (batch, …, #class) tensor of label probabilities (e.g. labels.sum(axis=-1) must be 1)

Returns

(batch, …) tensor of the cross-entropies for each entry.