src.spectral_clustering.initialize

src.spectral_clustering.initialize(X: torch.Tensor, num_clusters: int, method: str = 'plus') → numpy.array

Initialize cluster centers.

Parameters

• X – matrix

• num_clusters – number of clusters

• method – denotes different initialization strategies: 'plus' (default) or 'random'

Returns

initial state

Note

We support two initialization strategies: random initialization by setting method='random', or kmeans++ by setting method='plus'.