In the paper, online clustering secion, you mentioned to enforce equipartion, the vector are set to ones. As far as I understand, the sinkhorn algorithm is to find optimal transport between two distributions given a cost matrix.
Equipartition prevents the trivial solution where every clip has the same code. For more information regarding this (or the Sinkhorn-Knopp algorithm), please refer to:
Hello @hongluzhou ,
In the paper, online clustering secion, you mentioned to enforce equipartion, the vector are set to ones. As far as I understand, the sinkhorn algorithm is to find optimal transport between two distributions given a cost matrix.
My questions:
Can you please help?
cheers, xu