Neuron.fromWeights constructor

Neuron.fromWeights(int nin, { bool nonlin = true, })

Implementation

factory Neuron.fromWeights(int nin, {bool nonlin = true}) {
  // He Initialization for weights
  // For ReLU, standard deviation is sqrt(2 / nin)
  final double stdDev = math.sqrt(2 / nin);

  final w = List<Value>.generate(
      nin,
      (int index) => Value(
          // Sample from a normal distribution with mean 0 and stdDev
          // Dart's math.Random().nextDouble() gives a uniform distribution between 0.0 and 1.0.
          // A common way to approximate a normal distribution from a uniform one
          // for initialization is using Box-Muller transform or simpler approximations.
          // For practical purposes in this context, multiplying by stdDev
          // with a centered random value is a common heuristic if a true
          // normal distribution sampler isn't readily available or needed for simplicity.
          // A more rigorous approach would involve a proper Box-Muller implementation.
          (math.Random().nextDouble() * 2 - 1) *
              stdDev // centered random between -stdDev and +stdDev
          ),
      growable: false);

  Value b = Value(0.0); // Often good to initialize biases to 0
  return Neuron(ValueVector(w), b: b, nonlin: true);
}