1 paper
Shashank Raj, Kalyanmoy Deb
Adam and its variants dominate neural network training, but a single run only reveals whether a configuration works well after most of its budget is spent, a poor fit for hyperpara…