utilizing uncertainty estimation
[D] Utilizing Uncertainty Estimation to increase Auto-Label efficiency for Active Learning
I approved this post because even if you ignore the "Superb AI"-specific aspects, I think this post has value in providing an overview of uncertainty estimation methods. In other words, if you remove section 5, would this still be an interesting article? I do agree that it's somewhat disappointing that this blog post is structured as "here are the existing methods, which suck. Pay us to use our method instead!". However, I think that's reflected simply in the upvotes and general reaction. If the blog was instead, "here are the existing methods.