Industrial examples and research for automated, cognitive intelligence assistance for the Industy 4.0 and medical field. This research area combines AI-based analysis pipelines with distributed edge cloud continuum deployments. Click "View" for more interesting details.
comming soon
Anomaly Detection for Data Streams
Projects and Algorithms tackling the task of detecting abnormal behaviour within a high frequent data stream using unsupervised techniques.
Towards in-situ anomaly detection for edge clouds
Invited talk for the Sensor AI workshop 2020.
ATOS IT-challenge 2020 submission
Anomaly detection demonstration of out unsupervised online technique IFTM applied to cloud computing anomalies.
Unsupervsied Anomaly Detection - IFTM
Telecommunication system providers move their
IP multimedia subsystems to virtualized services in the cloud.
For such systems, dedicated hardware solutions provided a
reliability of 99.999% in the past. Although virtualization offers
more cost efficient usage of such services, it comes with higher
complexity for providing reliable running software components
due to the fragile computation stack. In order to hide the
impact of such problematic behaviors, automatic mechanisms
may help to detect degraded state anomalies in order to
execute remediation actions. This work introduces IFTM as a
framework for unsupervised anomaly detection in a distributed
environment based on real-time monitoring data. The proposed
approach consists of two key concepts using an automatic
identity function and threshold learning to distinguish between
normal and abnormal system behaviors. The evaluation is
performed on a testbed running an open source implementation
of the IP multimedia subsystem (Clearwater) executed on a
replicated Openstack cloud environment. Results show the
applicability of IFTM with high detection rates (98%) and
low number of false alarms.
IEEE Services 2018
Anomaly Detection for the Smart Factory
Cooperation project with Siemens aiming to integrate anomaly detection algorithms in to the industrial factories.
Anomaly Detection for Water Pumps
Cooperation project with KSB aiming to provide anomaly detection forecastings for water pumps for predictive maintenance.
Self-Healing Cloud
Within this cooperation project, we aim to create an immune system for computers. It should detect automatically anomalies, consider the root cause and select the appropriate counter actions to remediate the anomaly. Novel algorithms and testbeds are created together with our project partners Huawei Technologies and Deutsche Telekom.
Digital Education
Information about digital tools for visualization and simulating complex concepts for students.
We added further visual animations for memory placement strategies and resource usage.
SysprogInteract CPU-scheduling simulations
First CPU-scheduling visualization component for SysprogInteract. Please feel free to checkout the video and github repository.
TU Wimi+
TU Wimi+ is part of an BMBF project aiming to increase the quality of education. It consists of several smaller projects, which create new didactical media for including those into the education process. For the undergraduate computer science course Systemprogrammierung, we develop a novel simulation framework providing an interactive computer system simulation.
Project link from ZEWK and CIT.
Advanced Visualizations
Visualization is crutial for human understanding of complex data and analytics results. Next, you can see novel visualization techniques.
Interactive Decision Tree Visualization
Visualizing high dimensional decision tree models and data sets in the plane. Please feel free to checkout the video and github repository.
ViMEC Interactive Application for Micro-Cluster Visualizations
Digitalization increases the opportunity to collect vast amounts of data in a large scale manner. In order to handle the information overload, data mining techniques like online clustering are performed. A lot of online clusterers are based on the concept of micro-clusters in order to represent the given data stream. Based on its definition, micro-clusters can be represented as an n-sphere. Online clustering algorithms like BIRCH or DenStream use different strategies for maintaining the micro-clusters in evolving time series, but using the same underlying key concept storing a summarized version of the data stream in their models. We propose ViMEC, an application for multidimensional micro-cluster visualization, giving the user the opportunity to gain understanding of the internal behaviour of the clustering model. For a given time frame, ViMEC gives the user three different types of visualizations presenting different levels of details: Overview, Pair-view and Detail-view. These views combine not only a summary and detail representations for the different dimensions, but also aim to show different relations between dimensions. Preliminary results show, that large data sets with up to 20,000 data points can be visualized within less than 20 seconds.
Please feel free to checkout the video and paper.
Medical Decision Tree Visualization
Information overload in the medical domain is both visible
in the increased number of publications as well as in the
volume of patient data. In order to cope with this problem,
we propose a novel framework combining patient’s health
records with medical knowledge, which is based on medical
algorithms from frequently used guidelines. The framework
uses new types of animation and layout algorithms for visualizing
knowledge models in health records. At the Münster
University Hospital the framework is already in prototypical
use for education and communication purposes.
Please feel free to checkout the video and paper.
JavaD3
Java is not so nice when it comes to data visualizations. Javascript is instead very powerful for both creating visualizations, but also for creating advanced animations. JavaD3 provides a simple Java API to create visualizations like Timeseries plots, Histograms etc. In future, we are going to extend this library to create also animations. Checkout the github repository.
Big-Data Analytics
More information are provided about an EIT-Digital project MCloudDaaS and graph analytics.
EIT-Digital project MCloudDaaS
Multi-Cloud Data Analytics as a Service (MCloudDaaS) combines the benefits of Big Data Analytics and Multi-Cloud Systems.
Many companies struggle to find a convenient and affordable data analytic solution. Most cloud providers does not provide an easy to use cloud system Furthermore, clients have concerns like the lack of security, which leads to a mistrust against cloud services.
In order to overcome such barriers, MCloudDaaS provides innovative functionalities with regards to security, scalability and fault-tolerance. In addition, MCloudDaaS avoids a vendor lock-in of the customers, who would be in a traditional cloud system economically dependent of the provided DaaS solutions and therefore increasing costs when moving to a different cloud provider. To increase the trust of the customer, high security standards will be integrated. Moreover, MCloudDaaS provides rich management functionalities, predictions of performance and management for scalability. As a result, this leads to CAPEX reduction, because customers can switch to a cheaper provider.
Project partners: Atos, BT, CNR, TUB
Graph Processing with Grafungi
The simple graph library. It contains an embedded graph database, webservice for multi entity graphs and bayesian networks. Checkout the github repository.
Publications
Ferrer, A.J., Becker, S., Schmidt, F., Thamsen, L. and Kao, O., 2021. Towards a Cognitive Compute Continuum: An Architecture for Ad-Hoc Self-Managed Swarms. arXiv preprint arXiv:2103.06026. (accepted at Cloud2Things 2021)
Bogatinovski, J., Nedelkoski, S., Acker, A., Schmidt, F., Wittkopp, T., Becker, S., Cardoso, J. and Kao, O., 2021. Artificial Intelligence for IT Operations (AIOPS) Workshop White Paper. arXiv preprint arXiv:2101.06054.
Becker, Sören, Schmidt, Florian, Gulenko, Anton, Acker, Acker and Kao, Odej, 2020, December. Towards AIOps in Edge Computing Environments. In 2020 IEEE International Conference on Big Data (Big Data) (pp. 3470-3475). IEEE.
Styp-Rekowski, Kevin, Schmidt, Florian and Kao, Odej, 2020, December. Optimizing Convergence for Iterative Learning of ARIMA for Stationary Time Series. In 2020 IEEE International Conference on Big Data (Big Data) (pp. 2217-2222). IEEE.
Gulenko, A., Acker, A., Schmidt, F., Becker, S. and Kao, O., 2020, August. Bitflow: An In Situ Stream Processing Framework. In 2020 IEEE International Conference on Autonomic Computing and Self-Organizing Systems Companion (ACSOS-C) (pp. 182-187). IEEE.
Börner, A., Hübers, H.W., Kao, O., Schmidt, F., Becker, S., Denzler, J., Matolin, D., Haber, D., Lucia, S., Samek, W. and Triebel, R., 2020. Sensor Artificial Intelligence and its Application to Space Systems--A White Paper. arXiv preprint arXiv:2006.08368.
Anton Gulenko, Odej Kao, Florian Schmidt. Anomaly Detection and Levels of Automation for AI-Supported System Administration. Information Management and Big Data. SIMBig 2019. Communications in Computer and Information Science, vol 1070. Springer.
René Wetzig, Anton Gulenko, Florian Schmidt. Unsupervised Anomaly Alerting for IoT-Gateway Monitoring using Adaptive Thresholds and Half-Space Trees. IEEE International Conference on Internet of Things: Systems, Management and Security. IEEE, 2019.
Daniel Thilo Schroeder, Kevin Styp-Rekowski, Florian Schmidt, Alexander Acker and Odej Kao. Graph-based Feature Selection Filter Utilizing Maximal Cliques. In 2019 IEEE International Conference on Social Networks Analysis, Management and Security (SNAMS).
Florian Schmidt, Johannes Ohlemacher, Vincent Hennig. Case Study: Visualizing Computer System Programming Concepts for Education. SEFI 2019.
Florian Schmidt, Franz-Josef Schmitt, Laura Boeger, Arno Wilhelm-Weidner, and Nicole Torjus. Digital Teaching and Learning Projects in Engineering Education at Technische Universität Berlin. ASEE 2019.
Florian Schmidt, Florian Suri-Payer, Anton Gulenko, Marcel Wallschläger, Alexander Acker, and Odej Kao. Unsupervised Anomaly Event Detection for Cloud Monitoring using Online Arima. 11th IEEE/ACM International Conference on Utility and Cloud Computing (UCC 2018). IEEE, 2018.
Alexander Acker, Florian Schmidt, Anton Gulenko, and Odej Kao. Online Density Grid Pattern Analysis to Classify Anomalies in Cloud and NFV Systems. IEEE International Conference on Cloud Computing Technology and Science (CloudCom). IEEE, 2018.
Florian Schmidt, Florian Suri-Payer, Anton Gulenko, Marcel Wallschläger, Alexander Acker, and Odej Kao. Unsupervised Anomaly Event Detection for VNF Service Monitoring using Multivariate Online Arima. IEEE International Conference on Cloud Computing Technology and Science (CloudCom). IEEE, 2018.
Marcel Wallschläger, Anton Gulenko, Florian Schmidt, Alexander Acker and Odej Kao. Anomaly Detection for Black Box Services in Edge Clouds Using Packet Size Distribution. 2018 IEEE 7th International Conference on Cloud Networking (CloudNet). IEEE, 2018.
Dora Szücs and Florian Schmidt. Decision Tree Visualization for High-dimensional Numerical Data. In 2018 IEEE International Conference on Social Networks Analysis, Management and Security (SNAMS). IEEE, 2018.
Florian Schmidt, Anton Gulenko, Marcel Wallschläger, Alexander Acker, Vincent Kennig, Feng Liu, and Odej Kao. IFTM - unsupervised anomaly
detection for virtualized network function services. In 2018 IEEE International Conference on Web Services (ICWS). IEEE, 2018.
Anton Gulenko, Florian Schmidt, Alexander Acker, Marcel Wallschläger, Feng Liu, and Odej Kao. Detecting anomalous behavior of black-box services
modeled with distance-based online clustering. In 2018 IEEE International Conference on Cloud Computing (CLOUD). IEEE, 2018.
Florian Schmidt and Yannick Ehrenfeld. ViMEC: Interactive Application
for Micro-Cluster Visualizations. In
EuroVis 2018 - Posters. The Eurographics Association, 2018.
Marcel Wallschläger, Anton Gulenko, Florian Schmidt, Odej Kao, and Feng
Liu. Automated anomaly detection in virtualized services using deep packet
inspection. Procedia Computer Science, 110:510-515, 2017.
Alexander Acker, Florian Schmidt, Anton Gulenko, Reinhard Kietzmann,
and Odej Kao. Patient-individual morphological anomaly detection in
multi-lead electrocardiography data streams. In 2017 IEEE International
Conference on Big Data (Big Data), pages 3841-3846. IEEE, 2017.
Leon L Seifert, Anna Husing-Kabar, Florian Schmidt, Hauke Heinzow, and
Hartmut H Schmidt. Evaluation of potential risk factors and the role of
medication regimen in the development of hepatic encephalopathy after
transjugular intrahepatic portosystemic shunting in cirrhotic patients. In
HEPATOLOGY, volume 66, pages 277A-277A. WILEY 111 RIVER ST,
HOBOKEN 07030-5774, NJ USA, 2017.
Jannis Koch, Lauritz Thamsen, Florian Schmidt, and Odej Kao. Smipe:
Estimating the progress of recurring iterative distributed data
ows. In
2017 18th International Conference on Parallel and Distributed Computing,
Applications and Technologies (PDCAT), pages 156-163. IEEE, 2017.
Lauritz Thamsen, Benjamin Rabier, Florian Schmidt, Thomas Renner, and
Odej Kao. Scheduling recurring distributed data
ow jobs based on resource
utilization and interference. In 2017 IEEE International Congress on Big
Data (BigData Congress), pages 145-152. IEEE, 2017.
Florian Schmidt, Vincent Hennig, Sarah Köhler, Marcel Wallschläger, Anton
Gulenko, Hartmut Schmidt, and Odej Kao. Novel framework combining
health records with medical algorithms. In Proceedings of the 11th
EAI International Conference on Pervasive Computing Technologies for
Healthcare, pages 480-481. ACM, 2017.
Anton Gulenko, Marcel Wallschläger, Florian Schmidt, Odej Kao, and Feng
Liu. A system architecture for real-time anomaly detection in large-scale
nfv systems. Procedia Computer Science, 94:491-496, 2016.
Anton Gulenko, Marcel Wallschläger, Florian Schmidt, Odej Kao, and Feng
Liu. Evaluating machine learning algorithms for anomaly detection in
clouds. In 2016 IEEE International Conference on Big Data (Big Data),
pages 2716-2721. IEEE, 2016.
Lauritz Thamsen, Ilya Verbitskiy, Florian Schmidt, Thomas Renner, and
Odej Kao. Selecting resources for distributed data
ow systems according to
runtime targets. In 2016 IEEE 35th International Performance Computing
and Communications Conference (IPCCC), pages 1-8. IEEE, 2016.
Matthias Becker, Florian Schmidt, and Helena Szczerbicka. Applicability
of bio-inspired and graph-theoretic algorithms for the design of complex
fault-tolerant graphs. In 2013 IEEE International Conference on Systems,
Man, and Cybernetics (SMC), pages 2730-2734. IEEE, 2013.