New Polynomial-Augmented Neural Networks Enhance Function and PDE Approximation

Researchers have introduced Polynomial-Augmented Neural Networks (PANNs), a new architecture that merges deep neural networks (DNNs) with polynomial expansions. This hybrid approach aims to leverage the flexibility of DNNs for high-dimensional approximation and the rapid convergence of polynomials for smooth functions. The PANNs incorporate orthogonality constraints for stable training and accuracy, a basis pruning method to manage dimensionality, and a polynomial preconditioning strategy. Experiments show PANNs outperform standard DNNs in approximating smooth and non-smooth functions, as well as in solving partial differential equations. AI

IMPACT This novel architecture could improve the accuracy and efficiency of AI models in complex approximation tasks and scientific simulations.

RANK_REASON The cluster contains an academic paper detailing a new model architecture. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• deep neural networks
• Madison Cooley
• Polynomial-Augmented Neural Networks

• paper
• model release

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →