Data Mining and Machine Learning in Astronomy - Nicholas M. Ball & Robert J. Brunner

Because of the complex nature of galaxy morphology and the plethora of available approaches, a large number of further studies exist: Kelly & McKay [168] (Figure 1) demonstrate improvement over a simple split in u-r using mixture models, within a schema that incorporates morphology. Serra-Ricart et al. [169] use an encoder ANN to reduce the dimensionality of various datasets and perform several applications, including morphology. Adams & Woolley [170] use a committee of ANNs in a waterfall' arrangement, in which the output from one ANN formed the input to another which produces more detailed classes, improving their results. Molinari & Smareglia [171] use an SOM to identify E/S0 galaxies in clusters and measure their luminosity function. Genetic algorithms have been employed [173, 174] for attribute selection and to evolve ANNs to classify bent-double' galaxies in the FIRST [175] radio survey data. Radio morphology combines the compact nucleus of the radio galaxy and extremely long jets.