Name
Abstract | ||
|---|---|---|
In this work we derive an improvement of -APEX principal component analysis neural algorithms (8), based on a laterally-connected neural architecture, which arises from an optimization theory specialized for this topology. Such a class contains, as a special case, an APEX-like algorithm, but it also contains a subclass of algorithms that show interesting convergence features when compared with the original one. Data reduction techniques, as the Karhunen-Loeve Transform (KLT), aim at providing an efficient represen- tation of the data. The classical approach for evaluating the KLT requires the computation of the input data covariance matrix and then the application of a numerical procedure to extract the eigenvalues and the correspond- ing eigenvectors; reduction is obtained by projecting the data on the only eigenvectors associated with the most significant eigenvalues. When large data sets are handled, this approach is not practicable because the dimensions of the covariance matrix become too large to be manipulated. In order to overcome these problems, neural-network-based approaches were proposed. Neural Principal Com- ponent Analysis (PCA) is a second-order adaptive statistical data processing technique introduced by Oja (12) that helps to remove the second-order correlation among given random processes. In fact, consider the stationary multivariate random process and suppose its covariance matrix exists bounded. If is not diagonal, then the components of are statistically correlated. This second-order redundancy may be partially (or completely) removed by computing a linear operator such that the new ran- |
Year | Venue | DocType |
|---|---|---|
2000 | IJCNN (3) | Conference |
Citations | PageRank | References |
0 | 0.34 | 9 |
Authors | ||
3 | ||
Name | Order | Citations | PageRank |
|---|---|---|---|
| Simone Fiori | 1 | 494 | 52.86 |
| Saverio Costa | 2 | 0 | 0.34 |
| Pietro Burrascano | 3 | 21 | 4.77 |