Publikationen

The importance of data cleaning systems has continuously grown in recent years. Especially for real-time streaming applications, it is crucial, to identify and possibly remove anomalies in the data on the fly before further processing. The main challenge however lies in the construction of an appropriate data cleaning pipeline, which is complicated by the dynamic nature of streaming applications. To simplify this process and help data scientists to explore and understand the incoming data, we propose an interactive data cleaning system for streaming applications. In this paper, we list requirements for such a system and present our implementation to overcome the stated issues. Our demonstration shows, how a data cleaning pipeline can be interactively created, executed, and monitored at runtime. We also present several different tools, such as the automated advisor and the adaptive visualizer, that engage the user in the data cleaning process and help them understand the behavior of the pipeline.

A novel multimodal experimental setup and dyadic study protocol were designed to investigate the neurophysiological underpinnings of joint action through the synchronous acquisition of EEG, ECG, EMG, respiration and kinematic data from two individuals engaged in ecologic and naturalistic cooperative and competitive joint actions involving face-to-face real-time and real-space coordinated full body movements. Such studies are still missing because of difficulties encountered in recording reliable neurophysiological signals during gross body movements, in synchronizing multiple devices, and in defining suitable study protocols. The multimodal experimental setup includes the synchronous recording of EEG, ECG, EMG, respiration and kinematic signals of both individuals via two EEG amplifiers and a motion capture system that are synchronized via a single-board microcomputer and custom Python scripts. EEG is recorded using new dry sports electrode caps. The novel study protocol is designed to best exploit the multimodal data acquisitions. Table tennis is the dyadic motor task: it allows naturalistic and face-to-face interpersonal interactions, free in-time and in-space full body movement coordination, cooperative and competitive joint actions, and two task difficulty levels to mimic changing external conditions. Recording conditions - including minimum table tennis rally duration, sampling rate of kinematic data, total duration of neurophysiological recordings - were defined according to the requirements of a multilevel analytical approach including a neural level (hyperbrain functional connectivity, Graph Theoretical measures and Microstate analysis), a cognitive-behavioral level (integrated analysis of neural and kinematic data), and a social level (extending Network Physiology to neurophysiological data recorded from two interacting individuals). Four practical tests for table tennis skills were defined to select the study population, permitting to skill-match the dyad members and to form two groups of higher and lower skilled dyads to explore the influence of skill level on joint action performance. Psychometric instruments are included to assess personality traits and support interpretation of results. Studying joint action with our proposed protocol can advance the understanding of the neurophysiological mechanisms sustaining daily life joint actions and could help defining systems to predict cooperative or competitive behaviors before being overtly expressed, particularly useful in real-life contexts where social behavior is a main feature.

Future production technologies will comprise a multitude of systems whose core functionality is closely related to machine-learned models. Such systems require reliable components to ensure the safety of workers and their trust in the systems. The evaluation of the functional reliability and resilience of systems based on machine-learned models is generally challenging. For this purpose, appropriate test data must be available, which also includes abnormal cases. These abnormal cases can be unexpected usage scenarios, erroneous inputs, accidents during operation or even the failure of certain subcomponents. In this work, approaches to the model-based generation of an arbitrary abundance of data representing such abnormal cases are explored. Such computer-based generation requires domain-specific approaches, especially with respect to the nature and distribution of the data, protocols used, or domain-specific communication structures. In previous work, we found that different use cases impose different requirements on synthetic data, and the requirements in turn imply different generation methods [1]. Based on this, various use cases are identified and different methods for computer-based generation of realistic data, as well as for the quality assessment of such data, are explored. Ultimately we explore the use of Federated Learning (FL) to address data privacy and security challenges in Industrial Control Systems. FL enables local model training while keeping sensitive information decentralized and private to their owners. In detail, we investigate whether FL can benefit clients with limited knowledge by leveraging collaboratively trained models that aggregate client-specific knowledge distributions. We found that in such scenarios federated training results in a significant increase in classification accuracy by 31.3% compared to isolated local training. Furthermore, as we introduce Differential Privacy, the resulting model achieves on par accuracy of 99.62% to an idealized case where data is independent and identically distributed across clients.

There is a variety of software packages, toolboxes, or libraries for the analysis and processing of neurophysiological data such as EEG and MEG. Many of these solutions provide algorithms for both, sensor-space analysis and sourcespace analysis. Especially with the solutions that run on Windows machines, it is noticeable that the step of the volume model generation is usually not included, since the state-ofthe- art software for this (FreeSurfer) is a Unix-based software and thus not available forWindows machines. Therefore, our goal was to develop a fully-integrated software solution for Windows machines, accessing all processing steps already implemented in an existing toolbox and using FreeSurfer in the same system. Due to its widespread use, we chose MNE Python as the basis for our fully integrated software solution. We used the Windows Subsystem for Linux to create a virtual Linux kernel for the FreeSurfer installation. To demonstrate the workflow, the libeep, and AutoReject libraries have been added. A 64-channel EEG recording during right-hand movement (ME) and imagination (MI) was used to test the implemented workflow. The developed framework consists of several modules within Python, mainly using existing scripts and functions. The library libeep was integrated to read the EEG data with the ‘.cnt’, eeprope format. AutoReject was used to automatically interpolate detected bad channels or to reject complete epochs. FreeSurfer was successfully integrated and customized Python scripts enabled the communication between MNE Python on a Windows machine and FreeSurfer on a virtual Linux kernel. With the above-mentioned EEG dataset, we performed source reconstruction and were able to show ERD/S patterns for both, ME and MI. Our new, fullyintegrated software framework can be used on Windows machines to perform a complete process of source reconstruction.