Lifted Neural Networks

Armin Askari, Geoffrey Negiar, Rajiv Sambharya +1

We describe a novel family of models of multi- layer feedforward neural networks in which the activation functions are encoded via penalties in the training problem. Our approach i…