Sparse

reorder

spektral.utils.reorder(edge_index, edge_weight=None, edge_features=None)

Reorders edge_index, edge_weight, and edge_features according to the row-major ordering of edge_index.

Arguments

• edge_index: np.array of shape [n_edges, 2] (representing rows and columns of the adjacency matrix), indices to sort in row-major order.

• edge_weight: np.array or None, edge weight to sort according to the new order of the indices.

• edge_features: np.array or None, edge features to sort according to the new order of the indices.

Return

• edge_index: np.array, edge_index sorted in row-major order.
• edge_weight: If edge_weight is not None, edge_weight sorted according to the new order of the indices. Otherwise, None.
• edge_features: If edge_features is not None, edge_features sorted according to the new order of the indices. Otherwise, None.

---

sp_matrix_to_sp_tensor

spektral.utils.sp_matrix_to_sp_tensor(x)

Converts a Scipy sparse matrix to a SparseTensor. The indices of the output are reordered in the canonical row-major ordering, and duplicate entries are summed together (which is the default behaviour of Scipy).

Arguments

• x: a Scipy sparse matrix.

Return
A SparseTensor.