Publication accepted at the European Conference of Machine Learning and Principles and Practice of Knowledge Discovery in Databases
- Date: 18.06.2020
The paper “An uncertainty-based human-in-the-loop system for industrial tool wear analysis” of our master student Alexander Treiss, Jannis Walk and Niklas Kuehl was accepted for publication at The European Conference of Machine Learning and Principles and Practice of Knowledge Discovery in Databases. In the paper uncertainty measures based on Monte-Carlo dropout are used in the context of a human-in-the-loop system to increase the transparency and performance of the utilized convolutional neural network. A multiple linear regression is used to predict the quality of predictions. In case a prediction is estimated to be of low quality a human expert is asked to generate a prediction for the corresponding input. A simulation study demonstrates that the uncertainty-based human-in-the-loop system increases performance for different levels of human involvement in comparison to a random-based human-in-the-loop system. The system was developed in cooperation with Ceratizit Austria GmbH, in the research project Vertical Integration Analytics.
Feel free to contact the authors for more information.