std :: ml::nn
#include <std/ml/nn.h> source
#include <std/ml/nn.h>
#define OPT_ADAGRAD 6
per-weight lr / sqrt(sum of squared gradients) (Duchi et al. 2011)