Question for the implementation of feature aggregation.
peterzpy opened this issue · comments
peterzpy commented
Hi, I wonder why you implemented the feature aggregation using the new feature 'value', but not the original feature 'x' that calculated the similarity matrix. Since it may cause a collapse or anything else? Waiting for your response :)
Xu Ma commented
Hi, thanks for the question. There should be no big differences in terms of performance. I didn't see a particular reason to get a trivial solution or feature collapse.
peterzpy commented
Thanks for your reply.