Centers

[xiaokun-sh]

Hello, I was reading your SoftTriple Loss: Deep Metric Learning Without Triplet Sampling article and I wanted to ask you about an experiment you did in the article about the number of centres with or without a regulariser, and I wanted to ask how the number of centres with a regulariser dropped from an initialisation of 20 to 3,4 I would like to ask how the number of centres with a regulariser drops from the initial 20 to 3, 4