Neural networks encounter vanishing gradient problems as layers are stacked deeper, making training difficult to converge, and it was not until around 2010 that a series of technical breakthroughs solved the challenge of training deep networks, giving rise to the term deep neural network (DNN).