Bypassing Stationary Points in Training Deep Learning Models | IEEE Journals & Magazine | IEEE Xplore