This page collects the material for a 4-day course on the dissimilarity representation using the Matlab toolboxes DisTools and PRTools. (There is also a very short 2-hour set of examples). It is assumed that students are familiar with the concepts of pattern recognition and/or machine learning. The exercises described here are intended as an illustration of oral lectures and offer hand-on experiments by which a better understanding is gained. Moreover, the may serve as a start for further research and building applications.

Background material

Preparation

  • Download PRTools, DisTools and PRDisData.
  • Unzip and add them to your personal Matlab directory
  • Start Matlab
  • Change to your Matlab directory using the cd command
  • Add the above toolboxes to the Matlab path by (copy and paste)

addpath(fullfile(pwd,'prtools'))
addpath(fullfile(pwd,'distools'))
addpath(fullfile(pwd,'prdisdata')
)

  • Check the availability of PRTools, DisTools and PRDisData by

which ldc
which pe_em
which chickenpieces

They should all respond with a proper path.

Day 1: Introduction PRTools

  1. Introduction PRTools examples
  2. Scatterplots
  3. Datasets
  4. Datafiles
  5. Mappings
  6. Classifiers
  7. Evaluation
  8. Crossvalidation
  9. Learning curves
  10. Feature curves
  11. Dimension reduction
  12. Combining classifiers
  13. Dissimilarities

Day 2: Dissimilarity space

  1. Classifiers in dissimilarity space
  2. Dimension reduction
  3. Combining dissimilarity matrices
  4. Generalized dissimilarities
  5. Semi-supervised learning

Day 3: Embedding

  1. Pseudo-Euclidean-embedding
  2. Characterizations
  3. Euclidean corrections
  4. Classifiers in PE-space

Day 4: Applications

  1. Chickenpieces
  2. Brain MRI
  3. Spectra
  4. Music