Fuzzy HMC classifiers

Measuring effectiveness of a dynamic artificial neural network algorithm for classification problems

“…In this research we have only selected the simplest hierarchical version. Following research can examine DAN2 accuracy improvements using alternative approaches, such as those described in Kim et al (2003), Lee and Ersoy (2007) and Porter and Liu (1996).…”

“…Porter and Liu (1996) describe hierarchical classifiers as a subset of modular classifiers. They suggest that modular classifiers often arise when a combination of factors include a large number of classes, classes have difficult shapes (are not compact, convex, or connected), classes do not have distinct boundaries, boundaries are highly nonlinear, and misclassification of some points carries a high penalty.…”

“…Hierarchical neural networks may come in a variety of architectures. Both Porter and Liu (1996) and Lee and Ersoy (2007) describe a hierarchical neural network in which difficult training samples are rejected from the first layer and used for training in subsequent layers. Kim, Kehtarnavaz, Yeary, and Thornton (2003) first branched classes that separated easily, then used feature selection to build additional branches for the remaining classes, determining which features to use at each node of the tree.…”