A New Loss Function for Temperature Scaling to have Better Calibrated Deep Networks.
Azadeh Sadat MozafariHugo Siqueira GomesSteeven JannyChristian GagnéPublished in: CoRR (2018)
Keyphrases
- loss function
- pairwise
- learning to rank
- support vector
- logistic regression
- empirical risk
- reproducing kernel hilbert space
- loss minimization
- risk minimization
- boosting framework
- stochastic gradient descent
- hinge loss
- gradient boosting
- solution path
- weight vector
- machine learning
- regularization term
- multi class
- training data
- information retrieval