Geometry and Optimization of Shallow Polynomial Networks

Yossi Arjevani, Joan Bruna, Joe Kileel +2

We study shallow neural networks with monomial activations and output dimension one. The function space for these models can be identified with a set of symmetric tensors with boun…