scSLAT.model.train.check_align

scSLAT.model.train.check_align(embds, ground_truth, k=[5, 10], mode='cosine')

Check embedding correspondence in given distance (default cosine similarity) under ground truth

Parameters:

• embds (typing.List[torch.Tensor]) – List of graph features, each element is (node_num, feature_dim)

• ground_truth (torch.Tensor) – mapping ground_truth (2, node_num)

• k (typing.Optional[int]) – list of top k (only support 2 elements yet)

• mode (typing.Optional[str]) – distance quota

Return type:

typing.List[float]