[Seminar] SubSpace Methods for visual learning and recognition

On Friday December 5th, 2014  prof. Danijel Skocaj (Faculty of Computer and Information Science, University of Ljubljana) will give a seminar entitled ” Subspace methods”. On Friday December 12th, 2014  he will give another seminar entitled “Incremental interactive learning”.

05.12.2014 – 09.00-13.00 Subspace methods
12.12.2014 – 09.00-13.00 Incremental interactive learning

Venue: Aula Multimediale – Department of Mathematics and Computer Science (DIMI) – University of Udine – Via delle Scienze 206

Abstract: We often face high-dimensional data when modelling real world phenomena, especially when modelling visual data. However, usually these high-dimensional data lie in a much lower-dimensional space. Since low-dimensional data is significantly more convenient and efficient to handle, many methods exist that map high-dimensional data into a low-dimensional subspace, preserving certain properties of the original data. We will discuss several standard subspace methods, such as Principal component analysis, Linear Discriminant Analysis, and Canonical Correlation Analysis. We will derive these methods, and learn when and how to use them. We will also discuss how to extend the standard methods into incremental and robust versions that are more appropriate to use in real world scenarios.

Biosketch: Danijel Skocaj is an assistant professor at the University of Ljubljana, Faculty of Computer and Information Science. He is the vice-dean for research and the head of the Visual Cognitive Systems Laboratory. His research interests include artificial cognitive systems and intelligent robotics, computer and cognitive vision, as well as interactive social machine learning. He has published a number of papers and participated in or led several EU, national and industry funded projects addressing these research issues.


Organization and Contact: Dr. Christian Micheloni – christian.micheloni@uniud.it

Leave a Reply

Your email address will not be published. Required fields are marked *


This site uses Akismet to reduce spam. Learn how your comment data is processed.