Novel Fisher Subspace based Classifiers

In this book we present novel classifiers based on a new approach to estimate the Fisher Subspace. We face different classification tasks with the aim to stress our methods, to demonstrate their quality, and to show that these classifiers solve the following problems: 1. Manage unbalanced classes; 2. Solve the small sample size problem; 3. Deal with high dimensional data; 4. Manage classification tasks where the space dimensionality is approximately equal to the cardinality of the training set; 5. Relax the linear separability constraint reducing the overfitting problems.