Home | english  | Impressum | Sitemap | KIT

Applied and Algorithmic Challenges in Machine Learning

Applied and Algorithmic Challenges in Machine Learning
Typ: Proseminar (PS)
Semester: SS 2017
Zeit:

Kick-Off meeting

Freitag, 05.05.2017 von 08:00 - 11:30 Uhr
Raum 301 (3. Stock)
50.34 INFORMATIK, Kollegiengebäude am Fasanengarten 
 

LVNr.: 2400112
Hinweis:

Die Anmeldung erfolgt über das Sekretariat von Prof. Böhm, Geb 50.34 / Raum 367

Scheduling:

05.05.17: 08:00 - 11:30 (Room 301)
Kickoff - Presentation of offered topics and students choose from these.

10.05.17: 10:00 - 11:30 (Room 348)
Tutorial for technical necessities (Latex, GIT, etc. ) and deadline for seminar registration

24.05.17: 10:00 - 11:30 (Room 348)
Proposal presentation. Each student presents his initial and future work.

01.08.17: Submission of the slides for final presentations

07.08.17: 14:00 - 18:00 (Room 348) Final presentations

14.08.17: First draft of term paper

31.08.17 12:00 (commit to git repository) Final version of the term paper

Abstract:

Machine Learning is the foundation of modern data analytics as it allows for automated processing of data in regards to pattern recognition and discovery. With the rise of “Data Science” as a discipline, a large number of approaches and techniques are developed and refined for special applications. This seminar offers an introduction to some of these advanced methods.This seminar contains fundamental topics like Feature Selection and Feature Transformation or Generative Adversarial Networks., but also application-specific ones like the computation of operational risk using machine learning algorithms. In addition to that we give special consideration to streaming applications and their traits: While unbound input data and execution time requirements challenge static methods, the temporal component offers a semantically steady attribute across all applications. Other temporal related topics are in the field of text analysis. We specifically consider the analysis of the philosophical conceptual history, differentiate between the “word level” and the “text level” as well as giving rules about word usage and text structures that have been derived from large scale text analysis.

 

Topics

  • Methods to analyse philosophical conceptual history / J. Willkomm
  • Differences between 'word level' and 'text level' / J. Willkomm
  • Word/text rules derived from large scale text analysis / J. Willkomm
  • Feature Selection using MI on data streams (2) / M. Vollmer
  • Operational Risk Analysis Using Machine Learning / V. Arzamasov
  • Generative Adversarial Networks / G. Steinbuß
  • Feature Transformation on data streams / E. Fouché