Title "Enhancing Signal Processing Through Transparent AI in an Embedded Sensor"
10.04.2024, 14:15-14:45h
Session 1.5 Internet of Things / Computing
Abstract:
Traditional Machine Learning Algorithms often demand extensive training data, posing a significant challenge when dealing with sensor data from distributed EDGE IoT systems.
The author proposes a solution by introducing transparent rules rooted in human knowledge. This innovative approach not only enhances the accuracy of AI systems but also allows for continuous refinement by aligning rules with human judgments as a benchmark. Additionally, it empowers the system to provide clear explanations for its decisions. The author's system has been in operation for nearly two years, evaluating wastewater pumps servicing high-speed ICE trains for Deutsche Bahn.
Conference program: https://events.weka-fachmedien.de/embedded-world-conference/program
Comentários