Dynamic (Dis-)Information in Self-adapt...

This paper studies the effects of self-adaptive change of distributed search systems in which imperfectly informed search agents rather loosely collaborate and pursue objective functions which are not necessarily complements to each other. The results indicate that employing learning-based self-adaptation of major features of search systems may lead to high levels of systems’ performance, although the complexity of the search problem considerably tends to shape the effects of self-adaptation. The results further suggest that the selective effects of self-adaptation correspond to major features of the underlying search problem. This is of particular interest when the structure of the search problem is not known to the designer of the search system