PHP 클래스 MCordingley\Regression\Algorithm\GradientDescent\StoppingCriteria\GradientNorm
Good for Batch descent, when the gradient is fairly stable from one iteration
to the next. Not so good for Stochastic descent and MiniBatch with small
batch sizes, as those gradients won't settle near zero even as the weights
converge.
