input featuresweight vectors are optimized by maximizing the log-likelihooduse weight vectors → classifier by Bayes Decision Rule (argmax) → decision boundary