no code implementations • 9 Jan 2019 • W. Brent Daniel, Enoch Yeung
In this paper we presented a novel constructive approach for training deep neural networks using geometric approaches.