Immune Danger Zone Principle-Based Dynamic Learning Method
This chapter presents the immune danger zone principle-based dynamic learning model, which defines strategies for combining global learning with local learning based on the proposed multi-objective risk minimization principles . After an introduction and analysis of global learning and local learning, the necessity of building hybrid models is declared. As the foundation of the dynamic learning method, the proposed multi-objective learning principles are then stated. Afterward, strategies for combining global learning and local learning are described, as well as the analysis of local trade-off between capacity and locality. Implementations of the dynamic learning method are given. The relation of the dynamic learning method to multiple classifier combination is also analyzed. Finally, validation of the dynamic learning method is shown.