Uncertainty-aware Prediction Validator in Deep Learning Models for Cyber-physical System Data

The use of Deep learning in Cyber-Physical Systems (CPSs) is gaining popularity due to its ability bring intelligence CPS behaviors. However, both CPSs and deep have inherent uncertainty. Such uncertainty, if not handled adequately, can lead unsafe behavior. first step toward addressing such uncertainty quantify Hence, we propose a novel method called NIRVANA (uNcertaInty pRediction ValidAtor iN Ai) for prediction validation based on metrics. To this end, employ prediction-time Dropout-based Neural Networks models applied data. Second, quantified taken as the input predict wrong labels using support vector machine, with aim building highly discriminating validator model values. In addition, investigated relationship between quantification performance conducted experiments obtain optimal dropout ratios. We all four real-world datasets. Results show that negatively correlated Also, our ratio adjustment approach effective reducing correct predictions while increasing predictions.