Evaluation of multi-objective optimizers for cognitive radio using psychometric methods: Analysis using unidimensional and multidimensional Rasch models

Abstract

Item response models (IRMs) developed for use in fields such as education and psychology are applicable to cognitive radio testing due to parallels between cognitive radio and human cognition appear likely to enable efficient, and possibly adaptive testing of cognitive radios. A simulation study used unidimensional and multidimensional item response models to evaluate multi-objective cognitive engine optimizers based on two types of optimization algorithm: genetic algorithms and generalized pattern search. Data are presented in the context of cognitive radio and data are presented in a format that enables visualization of some characteristics of test items (optimization tasks) and optimizer performance identified by the IRMs. While the visualization provides intuitive confirmation of the IRM results, the IRMs identified additional significant effects that are not readily visible.