The documentation says:

The loss function to be used. Defaults to ‘hinge’, which gives a linear SVM. The ‘log’ loss gives logistic regression, a probabilistic classifier. ‘modified_huber’ is another smooth loss that brings tolerance to outliers as well as probability estimates.

When we use 'modified_huber' loss function, which classification algorithm is used? Is it SVM? If yes, how come it is able to give probability estimates, which is something it can't do with hinge loss?

  • $\begingroup$ The docs state that the modified_huber loss leads to calibrated responses that are converted to probabilities using an affine transformation (normalization). $\endgroup$
    – Emre
    Jul 6, 2017 at 18:08

1 Answer 1


The modified Huber loss is equivalent to a quadratically smoothed SVM with gamma = 2.

See also https://www.quora.com/What-algorithm-is-used-in-sklearn%E2%80%99s-SGDClassifier-when-a-modified-huber-loss-is-used/


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.