Comparison between generic SVM and 1NN detectors (SVMG, 1NNG) and a personalized 1NN detector (1NNP) in terms of AUC (mean and std).

For each person, the personalized 1NN is trained only with part of his or her own data, and tested with the remaining data. The generic SVM or 1NN in turn are trained with data from the remaining people but tested on the same validation set. This is repeated ten times for cross-validation.