Application of multiple self-organizing neutral networks: Flow pattern classification
Ye Mi, Lefteri H. Tsoukalas, M. Ishii · Transactions of the American Nuclear Society · 1997
For the purpose of horizontal-flow pattern classification, a multiple neural network system was developed with input from impedance-based measurement. After training the system, the tested result was in agreement with visual observation. A self-organizing neural network is a two-layer network that can cluster input data into several categories that include similar objects in the input data. The number of categories is specified subjectively and predetermined as the number of output nodes. The results of classification by the neural network can reveal the natural relations among the patterns in the input data.