-
|
When calling Line 642 in 0c99c8d I understand that this parameter controls whether you do recursive merging when assigning centroids using gvc: I notice that when discussing iterative clustering in your paper you note that recursive merging was used in the final step of clustering. When doing iterative clustering manually for large datasets, is this something that one should follow as well? How does this recursive clustering typically affect the consistency of the clusters? Do you typically end up with clusters that have many sequences that violate the id threshold against the representative/centroid? |
Beta Was this translation helpful? Give feedback.
Replies: 1 comment 1 reply
-
|
Strict gvc is false by default. This means that sometimes a member sequence will not satisfy the clustering criterion against its representative. It's not a big problem normally (see supplementary figs 5 and 6 of the Deepclust paper). You need to decide if strict adherence of the clustering criterion is important for your application. Without strict gvc, you get less clusters, more reduction of your data so better performance for the downstream application. |
Beta Was this translation helpful? Give feedback.
Strict gvc is false by default. This means that sometimes a member sequence will not satisfy the clustering criterion against its representative. It's not a big problem normally (see supplementary figs 5 and 6 of the Deepclust paper). You need to decide if strict adherence of the clustering criterion is important for your application. Without strict gvc, you get less clusters, more reduction of your data so better performance for the downstream application.