spectral_features_splitted

tangles.separations.finding.spectral_features_splitted(laplacian: spmatrix | ndarray, k: int, check_connected: bool = True, min_component_size_fraction: float = 0.01, ignore_small_components: bool = True) → ndarray

Compute spectral bipartitions of a graph after splitting into connected components.

Usually this is faster than the direct method spectral_bipartitions().

Parameters

laplaciansparse.spmatrix or np.ndarray

Laplacian matrix of a graph.

kint

Number of spectral bipartitions to compute.

check_connectedbool

Check if the graph is connected. If it is not connected, create bipartitions separating out the connected components first. If k is bigger than the number of connected components, the remaining bipartitions will split the components.

min_component_size_fractionfloat

Components of G smaller than min_component_size_fraction times the number of vertices are not analysed further.

ignore_small_componentsbool

If True, the small components are on the small side of all bipartitions. Otherwise there is a bipartition for every connected component.

Returns

np.array

-1/1-matrix containing the k spectral bipartitions of the graph as columns. The array might contain more than k separations, if the graph has more than k (big) connected components. The first few columns contain bipartitions separating out the connected components. If the graph is connected, the first column is constant 1.