Artificial Immune Systems: A Novel Approach to Electronic Nose Patterns Classification

摘要

This decade have seen the development of a novel pattern recognition paradigm, named Artificial Immune Systems, based on peculiar features expressed by mammalian immune systems. Apart from some controversy, typical of a newly developed paradigm, AISs semmes to express interesting features that could be exploited also in the field of artificial olfaction. In facts, AO practitioners have been exploring the neural metaphors since years now but have never explored the potentiality of the immune metaphor. In this paper, we experiment the use of AIRS, an AIS based classifier, comparing its performance with the well established FFNN classifier in a typical AO setting. Results are quite interesting highlighting the possibility to introduce AIS in artificial olfaction exploiting their novel features, particularly as regards as drift impact reduction and internal knowledge representation readability.