Abstract
We introduce methods for obtaining pretrained Geometric Neural Operators(GNPs) that can serve as basal foundation models for use in obtaining geometricfeatures. These can be used within data processing pipelines for machinelearning tasks and numerical methods. We show how our GNPs can be trained tolearn robust latent representations for the differential geometry ofpoint-clouds to provide estimates of metric, curvature, and other shape-relatedfeatures. We demonstrate how our pre-trained GNPs can be used (i) to estimatethe geometric properties of surfaces of arbitrary shape and topologies withrobustness in the presence of noise, (ii) to approximate solutions of geometricpartial differential equations (PDEs) on manifolds, and (iii) to solveequations for shape deformations such as curvature driven flows. We releasecodes and weights for using GNPs in the package geo_neural_op. This allows forincorporating our pre-trained GNPs as components for reuse within existing andnew data processing pipelines. The GNPs also can be used as part of numericalsolvers involving geometry or as part of methods for performing inference andother geometric tasks.