| Data sets consisting of a relatively small number of high-dimensional feature vectors often appear, e.g. in bioinformatics problems. This data structure complicates the design of classification or regression models. Complex layers of formal neurons (linear classifiers) can be designed on the basis of data sets composed of high-dimensional feature vectors. Linear classifiers of a given complex layer are designed on disjoint subsets of features obtained as a result of well-conditioned clustering. This feature clustering technique is related to matrix regularization. |
*** Title, author list and abstract as seen in the Camera-Ready version of the paper that was provided to Conference Committee. Small changes that may have occurred during processing by Springer may not appear in this window.