Interactive introduction to multi-way analysis in MATLAB
Next Chapter: Handling multi-way data




  1. Handling multi-way data in MATLAB
  2. Basic PARAFAC modeling
  3. Advanced PARAFAC modeling
  4. Multi-way calibration
  5. Advanced multi-way calibration
  6. More on preprocessing

  7. (No exercises)
  8. On missing values

  9. (No exercises)
This tutorial gives an introduction to multi-way analysis using MATLAB and the N-way Toolbox. The N-way Toolbox is compatible with MATLAB version 4.2 as well as MATLAB 5.x.

There are three main multi-way models: PARAFAC and Tucker for decompositions and N-PLS for calibration. N- PLS and Tucker are numerically stable and can hence be used as ordinary two-way methods without considering convergence etc. However, these models are not unique. The PARAFAC model is unique under mild conditions, but numerically it is sometimes unstable. Therefore one has to check convergence by doing multiple runs.

Otherwise the models can be interpreted and used very similar to two-way methods like PCA and PLS. The main obstacle is the confusion arising from handling the three-way data as they are stored in MATLAB as two-way matrices.

We appreciate all kinds of comments on this course: mis-spellings, suggestions for improvements, etc.

Download the toolbox. Uncompress the toolbox and be sure the path is included in your matlab path. Remember to download the data sets as well, since they are used throughout this course.

The material in this course and in the N-way Toolbox are given for free. However, we do appreciate appropriate referencing if you use our stuff. If you wish to refer to this WWW-course use the following:

R. Bro
The N-way on-line course on PARAFAC and PLS; 1998

Next Chapter: Handling multi-way data

The N-way tutorial
Copyright © 1998
R. Bro