Function sigmoidCrossEntropyWithLogits

tfjs-layers/src/losses.ts:171–191

(
    labels: Tensor, logits: Tensor)

 * @param logits The logits.
 */
export function sigmoidCrossEntropyWithLogits(
    labels: Tensor, logits: Tensor): Tensor {
  if (!util.arraysEqual(labels.shape, logits.shape)) {
    throw new ValueError(
        `logits and labels must have the same shape, but got shapes ` +
        `${JSON.stringify(labels.shape)} and ${JSON.stringify(logits.shape)}`);
  }
  return tidy(() => {
    // The logistic loss formula from above is
    //   x - x * z + log(1 + exp(-x))
    // For x < 0, a more numerically stable formula is
    //   -x * z + log(1 + exp(x))
    // Note that these two expressions can be combined into the following:
    //   max(x, 0) - x * z + log(1 + exp(-abs(x)))
    const reluLogits = tfc.relu(logits);
    const negAbsLogits = tfc.neg(tfc.abs(logits));
    return tfc.add(
        tfc.sub(reluLogits, tfc.mul(logits, labels)),
        tfc.log1p(tfc.exp(negAbsLogits)));
  });
}

export function binaryCrossentropy(yTrue: Tensor, yPred: Tensor): Tensor {
  return tidy(() => {