search

thuml / LogME

Code release for "LogME: Practical Assessment of Pre-trained Models for Transfer Learning" (ICML 2021) and Ranking and Tuning Pre-trained Models: A New Paradigm for Exploiting Model Hubs (JMLR 2022)
MIT License
203 stars 18 forks source link

Probabilistic meaning of LogME #25

Open ZiTingW opened 1 year ago

ZiTingW commented 1 year ago

Thank you for your code.

I am currently using LogME and have a question. I'm unsure whether the LogME is a probabilistic p(y|F), meaning if they are expected to fall within the [0, 1] range, or is a probabilistic density function with variables alpha and beta.

youkaichao commented 1 year ago

Hi, thanks for your interest. LogME is log p(y|F), but p(y|F) is a probabilistic density function since y is a continous variable. Therefore, it doesn't fall within the [0, 1] range.