Computational-Intelligence

Old News

Seminar Hot Topics in Neural Networks

Summerterm 2016

News

  • Die Slides on 'How to Review' are now available!
  • The presentation schedule has been actualized (version of 2016-06-03)

Overview

General

On this page, you find information on the seminar "Hot Topics in Neural Networks", which will be held during summer term 2016 by Christoph Doell at the Otto-von-Guericke University of Magdeburg. This site is actualized regularly during the semester.

Learning content of the seminar

The seminars purpose is to build and train key competences: scientific working, research, scientific writing, presenting and finalizing of a scientific topic.

The student activities consist of several parts: After the preparation of a scientific topic, it will be presented to all participants. Additionally a research paper on the chosen topic will be submitted as research paper. The submitted papers will then be peer-reviewed by the other participants. Possible flaws will be corrected and desired changes will be applied by the authors. The resulting finalized paper will be submitted again.

Technical content of the seminar

The technical content of the seminar are recent topics from the area of neural networks. As recent developments may use several complex parts, some of those build the early presentations. Possible Topics could be:

  • Neural Gas
  • Error Backpropagation: Variants, Alternatives
  • Recurrent Neural Networks: pros and cons
  • Hopfield Networks
  • Self-Organizing Maps
  • Deep Learning (Autoencoder, Restricted Boltzmann Machines)
  • Automatic Learning of Structures within Neural Networks
  • Applications (e.g.: real time translation, IQ-Tests, Learning to play Go)
  • Neural Networks in Hardware

Course

At the first appointment, each participant gets his own topic and correspondingly related literature. In the following weeks, one to three presentations are held, each followed by a short discussion. At this point participants shall actively give helpful feedback, in order to help the presenting students to improve their presenation skills. At about mid semester, the first version of the students papers are submitted. In the following weeks, these are reviewed giving the authors hints for an improvement. After applying them, the final version will be submitted.

Required knowledge

  • Basics in Computer Science
  • (helpful lectures) Neural Networks, Machine Learning

Organizational Rules for participation

The seminar is planned for up to 15 students in computer science and other related fields.

All participants are asked to register via FIN Registration Service.

For participation, the following efforts are needed:

  • successful oral presentation
  • 6-8 pages of an own research paper
  • reviewing of two other students' papers
  • Finalizing of the own research paper using the reviews
  • Participate in at least 90% of the presentations

For participating in the scientific seminar, we give 3 credit points. You can gain 6CP, by doubling the efforts e.g. writing two papers, holding two presentations, reviewing 4 papers, etc.

Appointments and rooms


WeekdayTimeRoomStart
Seminarthursday15.15-16.45G29-K05907.04.2016

Deadlines:

  • Submission of first version: 02.06.2016 (2016-06-02)
  • Submission of reviews: 23.06.2016 (2016-06-23)
  • Submission of final version: 07.07.2016 (2016-07-07)

Presentation schedule: Overview presentation schedule

Please submit your papers and reviews via easychair.

back to top of the page

Additional material

back to top of the page

Links to other websites

back to top of the page
en lang icon de lang icon Printable View - Recent Changes
Page last modified on June 23, 2016, at 11:50 AM by dockhorn