TY - JOUR A1 - Appelhans, Marie-Luise A1 - Kampmann, Matthias A1 - Mottok, Jürgen A1 - Riederer, Michael A1 - Nagl, Klaus A1 - Steffens, Oliver A1 - Dünnweber, Jan A1 - Wildgruber, Markus A1 - Roth, Julius A1 - Stadler, Timo A1 - Palm, Christoph A1 - Weiß, Martin Georg A1 - Rochholz, Sandra A1 - Bierl, Rudolf A1 - Gschossmann, Andreas A1 - Haug, Sonja A1 - Schmidbauer, Simon A1 - Koch, Anna A1 - Westner, Markus A1 - Bary, Benedikt von A1 - Ellermeier, Andreas A1 - Vögele, Daniel A1 - Maiwald, Frederik A1 - Hierl, Stefan A1 - Schlamp, Matthias A1 - Ehrlich, Ingo A1 - Siegl, Marco A1 - Hüntelmann, Sven A1 - Wildfeuer, Matthias A1 - Brückl, Oliver A1 - Sterner, Michael A1 - Hofrichter, Andreas A1 - Eckert, Fabian A1 - Bauer, Franz A1 - Dawoud, Belal A1 - Rabl, Hans-Peter A1 - Gamisch, Bernd A1 - Schmidt, Ottfried A1 - Heberl, Michael A1 - Thema, Martin A1 - Mayer, Ulrike A1 - Eller, Johannes A1 - Sippenauer, Thomas A1 - Adelt, Christian A1 - Haslbeck, Matthias A1 - Vogl, Bettina A1 - Mauerer, Wolfgang A1 - Ramsauer, Ralf A1 - Lohmann, Daniel A1 - Sax, Irmengard A1 - Gabor, Thomas A1 - Feld, Sebastian A1 - Linnhoff-Popien, Claudia A1 - Ławrowski, Robert Damian A1 - Langer, Christoph A1 - Schreiner, Rupert A1 - Sellmair, Josef ED - Baier, Wolfgang T1 - Forschung 2019 BT - Thema: Künstliche Intelligenz N2 - Bericht mit Forschungsprojekten aus verschiedenen Bereichen der OTH Regensburg mit dem Schwerpunktthema "Künstliche Intelligenz" und einem Gespräch zur "Medizin der Zukunft" T3 - Forschungsberichte der OTH Regensburg - 2019 KW - Forschung KW - Forschungsbericht KW - Künstliche Intelligenz Y1 - 2019 U6 - http://nbn-resolving.de/urn/resolver.pl?urn:nbn:de:bvb:898-opus4-7890 SN - 978-3-9818209-7-3 CY - Regensburg ER - TY - CHAP A1 - Gschossmann, Andreas A1 - Jobst, Simon A1 - Mottok, Jürgen A1 - Bierl, Rudolf T1 - A Measure of Confidence of Artificial Neural Network Classifiers T2 - ARCS Workshop 2019; 32nd International Conference on Architecture of Computing Systems, 20-21 May 2019, Copenhagen, Denmark N2 - Confidence in results of an Artificial Neural Networks (ANNs) is increased by preferring to reject data, that is not trustful, instead of risking a misclassification. For this purpose a model is proposed that is able to recognize data, which differs significantly from the training data, during inference. The proposed model observes all activations of the hidden layers, as well as input and output layers of an ANN in a grey-box view. To make ANNs more robust in safety critical applications, this model can be used to reject flawed data, that is suspected to decrease the accuracy of the model. If this information is logged during inference, it can be used to improve the model, by training it specifically with the missing information. An experiment on the MNIST dataset is conducted and its results are discussed. Y1 - 2019 UR - https://ieeexplore.ieee.org/document/8836211 SP - 153 EP - 157 PB - VDE-Verlag CY - Berlin ER -