The similarity matrix gets generated 3 times, once each from calling metacommunity_similarity, subcommunity_similarity, and normalized_similarity. As this is the order n-squared operation, we should avoid doing this unnecessarily. The use case is that the similarity matrix may be too large for RAM, so we don't want to cache the generated matrix; however, we could do all matrix multiplications in parallel, thus requiring the matrix only once. This should result in a nearly 3 times speed-up on large data sets.
The similarity matrix gets generated 3 times, once each from calling
metacommunity_similarity,subcommunity_similarity, andnormalized_similarity. As this is the order n-squared operation, we should avoid doing this unnecessarily. The use case is that the similarity matrix may be too large for RAM, so we don't want to cache the generated matrix; however, we could do all matrix multiplications in parallel, thus requiring the matrix only once. This should result in a nearly 3 times speed-up on large data sets.