On the Expressive Power of Deep Polynomial Neural Networks

Abstract

We study deep neural networks with polynomial activations, particularly their expressive power. For a fixed architecture and activation degree, a polynomial neural network defines an algebraic map from weights to polynomials. The image of this map is the functional space associated to the network, and it is an irreducible algebraic variety upon taking closure. This paper proposes the dimension of this variety as a precise measure of the expressive power of polynomial neural networks. We obtain several theoretical results regarding this dimension as a function of architecture, including an exact formula for high activation degrees, as well as upper and lower bounds on layer widths in order for deep polynomials networks to fill the ambient functional space. We also present computational evidence that it is profitable in terms of expressiveness for layer widths to increase monotonically and then decrease monotonically. Finally, we link our study to favorable optimization properties when training weights, and we draw intriguing connections with tensor and polynomial decompositions.

References in corpus (1)

• On the Expressive Efficiency of Sum Product Networks

Cited by in corpus (9)

• Deep Polynomial Neural Networks
• Depth-Width Trade-offs for ReLU Networks via Sharkovsky's Theorem
• On the Equivalence Between Temporal and Static Graph Representations for Observational Predictions
• Polynomial Neural Fields for Subband Decomposition and Manipulation
• Depth-Width Trade-offs for Neural Networks via Topological Entropy
• Identifiability for mixtures of centered Gaussians and sums of powers of quadratics
• PDE constraints on smooth hierarchical functions computed by neural networks
• Ladder Polynomial Neural Networks
• Linear independence of powers