Log loss
Log loss: A loss penalising probability assigned away from the correct class through the negative logarithm of correct confidence. Highly confident errors receive large penalties. It is sensitive to probabilistic quality and related to cross-entropy and maximum likelihood. This perspective connects the definition with practical behaviour in artificial intelligence systems.
In italiano: Log LossLegge V · Overfitting, bias e collasso dei modelliHighly confident errors receive large penalties. It is sensitive to probabilistic quality and related to cross-entropy and maximum likelihood.