Göm meny

Visual Representations for Machine Learning

Course goals

This course focuses on two specific topics in representation for machine learning in computer vision applications: Spectral clustering and Channel representations. Both of these representations will be described both from a theoretical point of view and applied to various problems.


Course participants are expected to have a good understanding of matrix decompositon (linear algebra, numerical methods).

Course organization

Preliminary, the course will consist of

  • 4 lectures, where theory and applications are presented. The first two on spectral clustering (Klas Nordberg) and the last two on channel representations (Michael Felsberg)
  • 2 seminars, course participants are expected to present excerpts from papers that describe applications and extensions of the lecture material.

The course gives 3hp to students who have attended all lectures and seminars, and have presented some amount of related material at both of the seminars.

Preliminary schedule

The course will be given during the end of the spring term 2015, preliminary no later than June 17. The schedule below might be subject to updates.

  • Lecture 1: Spectral clustering: introduction and confusion. May, 26, Thursday, 15.15 - 1700, Algoritmen! SLIDES
  • Lecture 2: Spectral clustering: from confusion to clarity. May, 30, Monday, 15.15 - 17.00. SLIDES
  • Lecture 3: Introduction to channel representations. June, 2, Thursday, 15.15 - 17.00. SLIDES
  • Lecture 4: Introduction to channel representations. June, 9, Thursday, 15.15 - 17.00. SLIDES
  • Seminar 1: spectral clustering. June, 15, Wednesday, 15.15 - 17.00. Gustav (Govindu), Abdo (Fowlkes et al.), Andreas (Zografos et al.)
  • Seminar 2: channel representations. June, 17, Friday, 10.15 - 12.00. Bertil (Suggested paper 1), Felix (Suggested paper 2), Hannes (Suggested paper 3).
All lectures, except the first one, and seminars will take place in 'Transformen'. (map), right above entrance 27.

Links and references

Contact information

  • Klas Nordberg, klas.nordberg@liu.se
  • Michael Felsberg, michael.felsberg@liu.se

Senast uppdaterad: 2016-06-09