softmax method

List<double> softmax(List<double> val)

Softmax activation function:

\text{Softmax}(x_{i}) = \frac{\exp(x_i)}{\sum_j \exp(x_j)} LaTeX in the absence of a better way to express it...

Implementation

List<double> softmax(List<double> val) {
  double max = mathutils.NEGATIVE_INFINITY;
  double sum = 0.0;
  val.asMap().forEach((i, v) => max = val[i] > max ? val[i] : max);
  val.asMap().forEach((i, v) => val[i] = exp(val[i] - max));
  val.asMap().forEach((i, v) => sum += val[i]);
  val.asMap().forEach((i, v) => val[i] /= sum);
  return val;
}