Fusion methods for unsupervised learning ensembles. (Q708016)

This book is a monograph on fusion methods used in unsupervised learning combined with ensemble learning. It covers neural nets with their learning strategies, especially training of topology preserving models. Chapter 3 discusses ensemble learning, including the most popular techniques of bagging, boosting, and mixture of experts. Chapter 4 presents the use of ensembles and unsupervised learning for outlier overcoming by generating graphical representations that are inspected by a naked eye. This way the user may recognize the most interesting clusters of data. Such projection methods are based on principal component analysis. The outliers may be recognized by using this technique combined with ensemble learning. Chapter 5 discusses combination of ensembles of topology-preserving maps known as fusion techniques. Chapter 6 presents the main contribution of the book: an improved version of the fusion algorithm. The discussed methodology was validated using many data sets, some artificial, some real-life.