search

wq2012 / SpectralCluster

Python re-implementation of the (constrained) spectral clustering algorithms used in Google's speaker diarization papers.
https://google.github.io/speaker-id/publications/LstmDiarization/
Apache License 2.0
513 stars 73 forks source link

Cluster Centers? #5

Closed mohammad-saber closed 5 years ago

mohammad-saber commented 5 years ago

Thanks for sharing your great code. I would like to know if I can get the center of clusters after prediction.

wq2012 commented 5 years ago

@mohammad-saber I think that's relatively easy to implement right?

Just use the prediction results to group the embeddings, and compute the embeddings for each group.

Question is how do you define center. It can be a Euclidean center, or the Riemann of the super unit sphere.