src.build_graphs.build_graphs

src.build_graphs.build_graphs(P_np: numpy.ndarray, n: int, n_pad: Optional[int] = None, edge_pad: Optional[int] = None, stg: str = 'fc', sym: bool = True, thre: int = 0) → Tuple[numpy.ndarray, numpy.ndarray, numpy.ndarray, int][source]

Build graph matrix \(\mathbf G, \mathbf H\) from point set \(\mathbf P\). This function supports only cpu operations in numpy. \(\mathbf G, \mathbf H\) are constructed from adjacency matrix \(\mathbf A\): \(\mathbf A = \mathbf G \cdot \mathbf H^\top\)

Parameters

  • P_np – \((n\times 2)\) point set containing point coordinates

  • n – number of exact points in the point set

  • n_pad – padded node length

  • edge_pad – padded edge length

  • stg – strategy to build graphs. Options: fc, near, tri

  • sym – True for a symmetric adjacency, False for half adjacency (A contains only the upper half)

  • thre – The threshold value of ‘near’ strategy

Returns

\(A\), \(G\), \(H\), edge_num

The possible options for stg:

'fc'(default): construct a fully-connected graph
'near': construct a fully-connected graph, but edges longer than ``thre`` are removed
'tri': apply Delaunay triangulation

An illustration of \(\mathbf G, \mathbf H\) with their connections to the graph, the adjacency matrix, the incident matrix is

../../_images/build_graphs_GH.png