296 followers
I don't disagree with the broader point, but the methods reportedly used to evaluate the models doesn't appear sufficient for the intended use case, and could lead to inflated performance scores. This is a common pitfall: www.nature.com/articles/s41... …