We present different topics of our research every Wednesday at 12:00 noon via zoom. For news about the EAD-Lunch talks and seminars please feel free to subscribe to EAD-Public@googlegroups.com. (Register here: https://groups.google.com/forum/?hl=de&fromgroups#!forum/ead-public)
contact: falko.gawantka(at)hszg.de
Explainability of artificial intelligence is an essential factor, which is required especially in critical domains e.g. legal requirements. The following research project focuses on an HR process, in which the quality and stability of several local XAI algorithms was tested. By increasing the number of repetitions of the algorithms, the results revealed that a more stable output with respect to the length of the confidence interval from feature importances could be achieved. In terms of the feature importance, by comparing the different algorithms, features importances could be strengthened/weakened and thus the respective relevances could be obtained.
contact: markus.wappler(at)hszg.de
Im Vortrag wird der probabilistische Aspekt des Quantencomputings erklärt. Davon ausgehend wird eine Methode entwickelt, gleichverteilte Listen zu erzeugen. Das eröffnet eine Möglichkeit, Nebenbedingungen für Optimierungsalgorithmen im Quantencomputer zu realisieren.
contact: falko.gawantka(at)hszg.de
There are many ways to explain black box models. This talk focuses on a game-theoretic approach. It is based on the Shapley Values, named after Lloyd Shapley. The approach of Shapley Values is introduced, i.e. the underlying computational process and the related problems are shown. Furthermore, the implementation by Lundberg and Lee (2017) is presented.