Skip to content

Morphology-Aware Neighbor Selection

Home / Open Questions / Morphology-Aware Neighbor Selection

Background: Interpolation of sparse, irregularly sampled physical simulation trajectories often relies on selecting a fixed number of neighboring data points in parameter space. Such fixed-neighbor approaches struggle to balance coverage and local accuracy, particularly when data points exhibit disparate morphologies or sensitive responses to initial conditions.

Question / Future Work: Developing dynamic or adaptive neighbor selection strategies that account for morphological similarity—rather than simple geometric distance—remains a critical hurdle in improving interpolation fidelity in high-dimensional physical parameter spaces. This includes handling variable neighbor counts and incorporating similarity-weighted barycentric schemes to avoid physical artifacts.

Why It Matters: This is a fundamental limitation in surrogate modeling and emulation for expensive scientific simulations, where sparse data often forces the inclusion of unrepresentative ‘neighboring’ samples.

Evidence: Future work could explore alternative weighting schemes that can accommodate variable numbers of neighbors or incorporate morphological similarity metrics into the neighbor selection process.

Metadata & Links