If I were the one asking you to predict these things, I would want your best subjective probability... especially if other agents might come to different conclusions.
<< modeling the loss (which is minus the loglikelihood) >>
I'm not aware of this loss function. But surely you can define other loss functions, no?
What is curious for me is that the proper scoring rule is not unique.
no subject
Date: 2010-11-19 04:10 pm (UTC)<< modeling the loss (which is minus the loglikelihood) >>
I'm not aware of this loss function. But surely you can define other loss functions, no?
What is curious for me is that the proper scoring rule is not unique.