HoloNet.predicting.adj_normalize#

HoloNet.predicting.adj_normalize(adj, cell_type_tensor, only_between_cell_type=True)#

Process the relationship among cells (such as multi-view CE network or spatial proximity matrix) to form the normalized adjancency matrix as the input of GNN.

Parameters
cell_type_tensor Tensor

cell_num * cell_type_num, derived from ‘get_continuous_cell_type_tensor’ or ‘get_one_hot_cell_type_tensor’ function.

adj Tensor

view_num * cell_num * cell_num, can be multi-view CE network or spatial proximity matrix

only_between_cell_type bool (default: True)

if True, will remove the edges between two cells with the same cell-type, to control the confounding factors.

Return type

Tensor

Returns

: Normalized adjancency matrix as the input of GNN.