K-core data filtering with infinity loop?

[rowedenny]

In the current implementation of data filter by interaction number, it iteratively drops the items and users based on the filter conditions.

Yet to my best knowledge, typically this dropout is only applied one-round instead of multiple round, otherwise, the volume of the original dataset will significantly decrease since one-round of drops will lead to some new "low frequent items and users"

Any reasons why we have an infinity loop here?

[hyp1231]

In RecBole, "k-core" has a similar meaning as "Degeneracy" in graph theory.

Many studies have followed this concept, one example is Amazon Review Data (2018).

K-cores (i.e., dense subsets): These data have been reduced to extract the k-core, such that each of the remaining users and items have k reviews each.

Thus, k is the number of interaction records of remaining users and items, and we have an infinity loop here.

[rowedenny]

I see, thanks!