Classification of Cm i energy levels using counterpropagation neural networks

Publisher Website
Google Scholar
Abstract
Two different types of counterpropagation neural networks are applied to the problem of classifying unknown Cm i energy levels. Four features—energy level, angular momentum, g factor, and isotope shift—are used to describe each level. One type of network is trained at the 100% level, while the other type is trained in excess of 96%. Performance on test sets is not as good, ranging from 81.2% to 93.7%. These results equal or surpass pattern recognition results obtained in an earlier study. Classifications for 12 odd-parity unknowns and 42 even-parity unknowns are also obtained and compared with the previous pattern recognition predictions.