✦ LIBER ✦
Feature extraction of polysaccharides by low-dimensional internal representation neural networks and infrared spectroscopy
✍ Scribed by Sven P. Jacobsson
- Publisher
- Elsevier Science
- Year
- 1994
- Tongue
- English
- Weight
- 729 KB
- Volume
- 291
- Category
- Article
- ISSN
- 0003-2670
No coin nor oath required. For personal study only.
✦ Synopsis
A new method for exploratory data analysis of spectroscopic data by neural networks is described. The method is based on the weight distribution associated with objects in narrow layered neural networks in which the input spectra are identical to the output spectra. The objects are displayed in 2-or 3-dimensional plots in analogy to principal component plots. The information content of the plots generated by low-dimensional internal representation neural networks, in their non-linear mode, appears to be at least as good as that of principal component analysis.