AD
ID:611170311
what loss functoins may be used for binary classificaton? There are only 2 of them in link
https://catboost.ai/docs/concepts/loss-functions-classification.html
I think about minimum probability to mistake for not balanced data,
for example probaility of 1s is 0.05 and probability of 0s is 0.95, so we need F1 metric as loss function?
https://scikit-learn.org/stable/modules/generated/sklearn.metrics.f1_score.html with 'weighted' and maybe micro
loss function should be the same as metric? why to optimise for one function and measure performance by anothe function
as he mentioned logloss is useless https://youtu.be/xl1fwCza9C8?t=1383
https://catboost.ai/docs/concepts/loss-functions-classification.html
I think about minimum probability to mistake for not balanced data,
for example probaility of 1s is 0.05 and probability of 0s is 0.95, so we need F1 metric as loss function?
https://scikit-learn.org/stable/modules/generated/sklearn.metrics.f1_score.html with 'weighted' and maybe micro
loss function should be the same as metric? why to optimise for one function and measure performance by anothe function
as he mentioned logloss is useless https://youtu.be/xl1fwCza9C8?t=1383
Вроде на все вопросы выше в ods чате и в issues-ах ответила, как могла, наверное тут третий раз не имеет смысла дублировать.
