Creating exhaustive spatial region in design space

1 次查看（过去 30 天）
Maaz Ahmad
Maaz Ahmad 2020-9-4
I am working with high-dimensional input data of the order of 20, and having nearly 4000 sample points. The response dimension is 1. I want to create spatial regions in my domain space but it should be exhaustive and non-overlapping. Delaunay triangulations look like a perfect solution but its just not possible to construct them in high-dimensions due to storage issues. I thought of constructing large grids too but again, even if I have 2 bins per dimension, I end up with 2^20 grids for 20-dimensional data. Is there any alternative to address this curse of dimensionality so that I can obtain regions computationally efficiently. In the worst case, I may even consider overlapping geometries, and/or non-exhaustive design.
By exhaustive, I mean each sample point is part of some geometrical region.

请先登录，再回答此问题。

回答（0 个）

类别

在 Help Center 和 File Exchange 中查找有关 Delaunay Triangulation 的更多信息

标签

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by