News from the Artificial Intelligence Group

The chair of artificial intelligence deals with the wide field of machine learning. In particular the chair concentrates on the development and implementation of learning algorithms that solve challenging problems.

Lecture "Große Daten, Kleine Geräte" ("Big Data, Small Devices") in the Science Notes

Science Notes Poster

Intelligent fabrics, fitness wristbands, smartphones, cars, factories, and large scientific experiments are recording tremendous data streams. Machine Learning can harness these masses of data, but storing, communicating, and analysing them spends lots of energy. Therefore, small devices should send less, but more meaningful data to a central processor where additional analyses are performed.

(more...  )

Katharina Morik among the leaders of Germany's "Platform Learning Systems"

WG leaders

Germany ranks among the pioneers in the field of learning systems and Artificial Intelligence. The aim of the Plattform Lernende Systeme initiated by the Federal Ministry of Education and Research is to promote the shaping of Learning Systems for the benefit of individuals, society and the economy. Learning Systems will improve people’s quality of life, strengthen good work performance, secure growth and prosperity and promote the sustainability of the economy, transport systems and energy supply.

(more...  )

Merry Christmas and a Happy New Year 2018

Christmas 2017

We wish you a merry christmas and a happy new year.

We used the Deep Visualization Toolbox of Yosinski for creating a nice picture of the visit of the three holy kings.

Best Paper Award of the International Conference on Spatial Information Theory (COSIT) 2017

Best Paper Award of the International Conference on Spatial Information Theory (COSIT) 2017

The joint work "On Avoiding Traffic Jams with Dynamic Self-Organizing Trip Planning" of Thomas Liebig and Maurice Sotzny received the Best Paper Award of the International Conference on Spatial Information Theory (COSIT) 2017.

Vacant Professorship(W2) Position

TU Dortmund University is seeking an outstanding scientist in the field of data mining of large datasets with a current research perspective and publications in high-ranked international venues. Applicants should complement the research activities of the Faculty for computer science and contribute to interdisciplinary collaborative research projects, especially the collaborative research centre CRC 876 “Providing Information by Resource-Constrained Data Analysis“.

Further information is given in the linked document

Introduction to Machine Learning for Users and the General Public

The Academy of Engineering has presented an online course on machine learning at CeBIT: http://www.acatech.de/de/projekte/projekte/mooc-maschinelles-lernen.html

After an overview presented by Prof. Dr. Stefan Wrobel (Fraunhofer St. Augustin), Katharina Morik introduces two basic methods with application examples from her many years of practical experience: the support vector machine (SVM) and decision trees. Kristian Kersting presents probabilistic graphical models.

Klassifikation und Regression - Stützvektormethode (Classification and Regression - SVM)

Download 120 MB [mp4]
Source: acatech

Klassifikation und Regression - Entscheidungsbäume (Classification and Regression - Decision Trees)

Download 87 MB [mp4]
Source: acatech

Probabilistische Graphische Modelle (Probabilistic Graphical Models)

Download 86 MB [mp4]
Source: acatech

Outstanding Graduates of TU Dortmund receive Hans-Uhde Award

Award Recipients

Four graduates of TU Dortmund received the Hans-Uhde Award for their outstanding theses. Niklas Haarmann (Faculty of Bio- and Chemical Engineering), Chris Kittle (Faculty of Electrical Engineering and Information Technology) and Lukas Pfahler (Faculty of Computer Science) achieved a master's degree and graduated as valedictorians. Christian Gehring (Faculty of Mechanical Engineering) received a grade of 1,0 for his bachelor's thesis. Additionally, three graduates of FH Dortmund and one employee of Uhde Inventa-Fischer GmbH were decorated by the Hans-Uhde Foundation.

The graduates of TU Dortmund were awarded a golden coin, a certificate and a monetary price by Guido Baranowsky, chairman of the Hans-Uhde foundation. In his thesis, Lukas Pfahler explored the question how to enable computers to learn German grammar. The ceremony took place at thyssenkrupp Industion Solutions AG in Dortmund. The ceremonial address — "Precision Medicine and Foundational Research; Innovation with Potential" — was delivered by Prof. Daniel Rauh. The goal of the Hans-Uhde Foundation is to promote Science, Schooling and Education. This is why it annually decorates outstanding students as well as pupuils. The award ceremony was attended by both Hans and Roswitha Uhde until 2011, when Friederich Uhde passed. The widowed Roswitha Uhde continued to attend the ceremonies up until her passing in 2017.


Hans-Uhde Award 2017

Lukas Pfahler, M.Sc.
TU Dortmund, Faculty of Computer Science
Master's Thesis: Explicit and Implicit Feature Maps for Structured Output Prediction


Marco Stolpe Defends His Dissertation at LS8

Marco Stolpes Disputation

Marco Stolpe has successfully defended his dissertation “Distributed Analysis of Vertically Partitioned Sensor Measurements under Communication Constraints”. His thesis was supervised by Katharina Morik and can be summarized (in German) as follows:

Schwerpunkt der Arbeit ist die verteilte Analyse großer Mengen vertikal partitionierter Sensordaten unter Berücksichtigung von Kommunikationsbeschränkungen. Hierbei hängt die vorherzusagende Zielgröße jeweils von an unterschiedlichen Knoten im Netzwerk gespeicherten Merkmalswerten ab. Das Szenario hat vielfältige Anwendungen im Kontext des Internet of Things und Industrie 4.0, wie etwa die Vorhersage der finalen Produktqualität anhand von an verschiedenen Bearbeitungsstationen erfassten Prozessparametern, die Vorhersage des Gesamtstromverbrauchs anhand des zuvor erfassten Verhaltens unterschiedlicher Stromabnehmer im Smart Grid oder die Vorhersage von Verkehrsflüssen in Smart Cities. Das Szenario erweist sich als besonders herausfordernd in Fällen, in denen Kommunikation oder Energie zu beschränkt sind, um alle Daten zu zentralisieren, da bereits für die Vorhersage Daten unterschiedlicher Knoten zusammengeführt werden müssen. In der Dissertation werden, motiviert durch eine Fallstudie zur Qualitätsvorhersage in verketteten Produktionsprozessen in der Stahlindustrie, kommunikationseffiziente Algorithmen für drei unterschiedliche Problemstellungen der verteilten Datenanalyse entwickelt: (1) Die lokale Reduktion von Messwerten unmittelbar dort, wo sie erfasst werden (also noch vor ihrer Übertragung), (2) die Reduktion von Messwerten, die zwischen lokalen Knoten und einem zentralen Koordinator übertragen werden und (3) die Reduktion von Informationen über vorherzusagende Zielgrößen, die zwischen Knoten übertragen werden. Die Algorithmen reduzieren die übertragene Datenmenge im Vergleich zur Übermittlung aller Daten in einem Netzwerk jeweils um ca. eine Größenordnung, bei ähnlicher Vorhersagegüte. Algorithmus (3) basiert wiederum auf einem neu entwickelten Ansatz für das relativ neuartige Problem des Lernens aus Label-Verhältnissen, dessen Lösung weitere Anwendungen im Kontext von Industrie 4.0 erschließt.


Christian Pölitz Defends His Dissertation at LS8

Christian Plitzs Disputation

Christian Pölitz has successfully defended his dissertation “Automatic Methods to Extract Latent Meanings in Large Text Corpora”. His thesis was supervised by Katharina Morik and can be summarized as follows:

This thesis concentrates on Data Mining in Corpus Linguistic. We show the use of modern Data Mining by developing efficient and effective methods for research and teaching in Corpus Linguistics in the fields of lexicography and semantics. Modern language resources as they are provided by Common Language Resources and Technology Infrastructure (http://clarin.eu) offer a large number of heterogeneous information resources of written language. Besides large text corpora, additional information about the sources or publication date of the documents from the corpora are available. Further, information about words from dictionaries or WordNets offer prior information of the word distributions. Starting with pre-studies in lexicography and semantics with large text corpora, we investigate the use of latent variable methods to extract hidden concepts in large text collections. We show that these hidden concepts correspond to meanings of words and subjects in text collections. This motivates an investigation of latent variable methods for large corpora to support linguistic research.

Show news archive