Fast Methods for Reducing Dimensionality of Spectral Data for Their Visualization

We have developed fast methods for reducing the dimensionality of spectral data, such as IR spectroscopy data, chromatography data, etc. In contrast to the widely known methods for projection of data with dimensionality N read counts onto spaces of lower dimensionality, having computational complexi...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:Journal of applied spectroscopy 2019-03, Vol.86 (1), p.101-105
Hauptverfasser: Vagin, V. A., Krasnov, A. E., Nicol’skii, D. N.
Format: Artikel
Sprache:eng
Schlagworte:
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:We have developed fast methods for reducing the dimensionality of spectral data, such as IR spectroscopy data, chromatography data, etc. In contrast to the widely known methods for projection of data with dimensionality N read counts onto spaces of lower dimensionality, having computational complexity of order N × N , proportional to the dimensionality of the covariance matrices for the data, in order to reduce the computational time we propose using new methods that can be realized in a sliding window over n read counts. As a result, the fast methods have computational complexity of order n × N . We present the results of computer experiments on reducing the dimensionality of IR spectra for automotive gasolines. The problem of reducing the dimensionality of IR spectra is important for both their graphical visualization and decreasing the multicollinearity and reducing the influence of noise in simulating behavior or analysis of parameters depending on the spectral characteristics.
ISSN:0021-9037
1573-8647
DOI:10.1007/s10812-019-00788-3