search

markus93 / NN_calibration

Calibration of Convolutional Neural Networks
MIT License
157 stars 41 forks source link

NLL in Temperature scaling #10

Closed dsvrsec closed 3 years ago

dsvrsec commented 3 years ago

In temperature scaling code,negative log likelihood is considered not considered as loss function where as only Log likelihood is considered.Is there any reason or is it an error

markus93 commented 3 years ago

Hey!

Actually, it is the same thing, the log loss is implemented as negative log-likelihood of a logistic model. More information in documentation of sklearn in here.