squidpy.gr.neighbors.DelaunayBuilder

class squidpy.gr.neighbors.DelaunayBuilder(radius=None, transform=None, set_diag=False, percentile=None)[source]

Build a generic point-cloud graph from a Delaunay triangulation.

Delaunay triangulation connects observations into triangles such that no other observation lies inside the circumcircle of each triangle. Unlike GridBuilder(delaunay=True), this builder uses geometry-based connectivity and stores real Euclidean edge lengths.

radius only controls post-construction edge pruning; the triangulation itself is unchanged. A tuple (min, max) keeps edges with Euclidean length in that interval. A scalar r is shorthand for (0.0, r), i.e. keep edges with length at most r. None keeps every edge.

See spatial_neighbors_delaunay() for the user-facing API or spatial_neighbors_from_builder() for direct builder usage.

Methods

`build`(coords)
`build_graph`(coords)	Construct raw adjacency and distance matrices.
`combine`(mats, ixs)	Combine per-library results into a single graph.
`postprocessors`()	Return post-build processing steps for `(adj, dst)`.
`uns_params`()	Parameters stored in `anndata.AnnData.uns` after graph construction.