Low-Rank Neural Regression Models for Spatial Data on Complex Domains

A methodological paper proposes models for spatial data on non-Euclidean domains like manifolds and graphs. The 1.4 MB resource includes PDF and MD files authored by Luigi Ippoliti and published under a CC-BY-4.0 license on figshare in May 2026. It demonstrates the framework through four applications with increasing domain complexity.

Use Cases

• Developing spatial regression models based on spectral representations using eigenfunctions of the graph Laplacian.
• Modeling random functions over complex geometries based on the described low-rank spatial regression framework.
• Applying feed-forward neural networks to spatial data based on the extension to non-linear regression for non-Euclidean domains.
• Analyzing data with locally varying dependencies based on models that explicitly account for underlying domain geometry.

Strengths

• The description details a specific methodological framework for complex spatial data.
• The resource is published under a permissive CC-BY-4.0 license.
• The approach is illustrated through four distinct applications mentioned in the description.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• The primary files are PDF and MD formats, suggesting the resource is documentation rather than a structured dataset.

Provenance

Source

figshare

Freshness

Last updated 2026-05-11 12:37:52; freshness should be verified.