Researchers have developed a new framework for creating scalable and flexible covariance kernels for Gaussian processes (GPs). This method directly learns the covariance structure using deep neural architectures and a regression-type parameterization derived from Vecchia approximations. The approach leverages permutation-preserving functions, inspired by the permutation-equivariant structure in Vecchia factorization, to enhance training stability and data efficiency. AI
Summary written by gemini-2.5-flash-lite from 2 sources. How we write summaries →
IMPACT Introduces a novel framework for learning covariance structures in Gaussian processes, potentially improving scalability and data efficiency in machine learning applications.
RANK_REASON The cluster contains an arXiv preprint detailing a novel framework for Gaussian processes using deep neural networks.