Title :
Unsupervised classification by spectral ICA
Author_Institution :
Office of Naval Res., Arlington, VA, USA
Abstract :
Unsupervised classification is defined such that the information required to do so must be learned and derived directly and solely from the data alone, this is consistent with the classical definition "unlabelled data" ATR by Duda and Hart. Such a truly unsupervised methodology is presented for space-variant imaging for breast cancer detection by means of a spectral-ICA methodology rather than by spatial-ICA for space-invariant imaging.
Keywords :
biomedical optical imaging; image classification; independent component analysis; infrared imaging; mammography; medical image processing; unsupervised learning; breast cancer detection; hyperspectral sensors; independent component analysis; infrared radiation; neural nets; space-variant imaging; spectral ICA; thermal breast scanning; unsupervised classification; Breast cancer; Cancer detection; Independent component analysis; Lagrangian functions; Optical imaging; Physics; Pixel; Remote sensing; Surveillance; Thermodynamics;
Conference_Titel :
Neural Information Processing, 2002. ICONIP '02. Proceedings of the 9th International Conference on
Print_ISBN :
981-04-7524-1
DOI :
10.1109/ICONIP.2002.1198976
