Figure 7.

The verification process for adaptive learning classifiers. Samples with the same concept compose a session. Sessions with different concepts compose a session sequence. Samples in the session sequence are fed to the updating cycle one by one, and recognition accuracy (RA_i) is concluded in the unit of session. During an updating cycle, we firstly use the predictive model (PM) to classify the sample and make the prediction, then use the sample and its corresponding supervised label information (in supervised adaptive learning scenarios) or prediction result (in unsupervised adaptive learning scenarios) to retrain the predictive model.