Pattern recognition, reflections from a chemometric point of view

Cluster analysis and projections on latent variables are two classical paradigms of pattern recognition in chemometrics. The former describes the world in terms of distances (or dissimilarities) between its various elements. It can be considered as a descriptive (or Aristotelian) view of the world which produces taxonomies of its elements. The latter considers fundamental properties of the elements, that are often hidden from direct observation. It constitutes an explanatory (or Platonic) interpretation of the world which unveils simple structures that lie behind it. phor for pattern recognition is developed which is based on the property of electronic and mechanical systems to resonate at their eigenfrequency when stimulated by a mixture of frequencies. The latent vectors produced from tabulated data by means of singular vector decomposition and partial least squares can be assimilated with the eigenvibrations of such resonating systems. Latent vectors provide a link between the two paradigms of pattern recognition which were described above. Indeed, latent vectors extracted from a data table are the same as those derived from the corresponding distance matrix (after transformation of the latter into a variance—covariance matrix). scientific, philosophical and esthetical perspective, one may hope that chemometrics, apart from solving important problems of immediate interest, will also be capable of producing a novel and persisting view of the world, in the same way as has been achieved in biometrics and in psychometrics.