Rough-fuzzy-neurocomputing based on rough mereological calculus of granules¹

Issue title: Hybrid Intelligence using rough sets

Article type: Research Article

Authors: Polkowski, Lech^{a; b; *}

Affiliations: [a] Polish-Japanese Institute of Information Technology, Warsaw, Poland | [b] Department of Mathematics and Computer Science, University of Warmia and Mazury, Olsztyn, Poland

Correspondence: [*] Address for correspondence: PJIIT, Koszykowa 86, 02008 Warsaw, Poland. Tel.: +48 2258 44678; Fax: +48 2258 44501; E-mail: [email protected]

Note: [1] This is an extended and updated version of a talk at HIS’03, Melbourne, Australia.

Abstract: In this paper, we propose a scheme for merging rough, fuzzy, and cognitive (neural) computation methods. The underlying paradigm is that of rough mereology. Rough inclusions that are logical connectives of rough mereology allow for constructing granules of knowledge that constitute elementary objects for calculi merging rough, fuzzy and neural schemes. The presented scheme is based on mereological calculus of granules due to the author. The paper is intended as a research paper yet its aim is also to acquaint the reader with intuitions and motivations essential to the proposed approach. The idea that permeates the paper is that rough inclusions provide a bridge between rough and fuzzy theories linking them across the gap resulting from distinct approaches at uncertainty of knowledge.

Keywords: rough-neural computing, rough mereology, rough sets, reasoning under uncertainty

DOI: 10.3233/HIS-2005-2202

Journal: International Journal of Hybrid Intelligent Systems, vol. 2, no. 2, pp. 91-108, 2005