10th VIPS Advanced school on Computer Vision and Pattern Recognition

Verona 23-26 Semptember 2013

Dissimilarity-based Representation for Pattern Recognition

Sept 23  (10,00 : 12,00 -- 13,30: 16,30) : Representations for Pattern Recognition

Representation and generalization;

compactness, true representations;

vector spaces and structural representations;

features, pixels and dissimilarities;

densities, domains and distances;

short review of classification and evaluation.

Sept 24  (10,00 : 12,00 -- 13,30: 16,30) : The dissimilarity space

properties of distances and dissimilarities;

features versus dissimilarities;

definition and properties of the dissimilarity space;

classification and cluster analysis in the dissimilarity space;

prototype selection versus feature selection.

Sept 25  (10,00 : 12,00 -- 13,30: 16,30) : Pseudo-Euclidean embedding

embedding; multi-dimensional scaling and kernel embedding;

pseudo-Euclidean (PE) space;

correction procedures;

classification in PE space;

embedding of new objects;

transductive learning;

relation with the dissimilarity space;

causes of non-Euclidean dissimilarities, are they informative?

Sept 26 (10,00 : 12,00 -- 13,30: 16,30) : Applications

Images, time signals and spectra, graphs and sequences;

focus of histograms and spectra;

distance measures; relation with metric learning;

examples from computer vision, medical imaging,

hyperspectral imaging, seismics and/or chemometrics.

Here are some links with useful materials:

- helpful documentation here

- lab examples and experiments here

- In the blog on Pattern Recognition Tools regularly topics related to this course have   been discussed, e.g. start at 22 October 2012. See here

Here are instructions for final exam !!