Fuzzy Rule Interpolation Matlab Toolbox - FRI Toolbox


Team Members:

Zsolt Csaba Johanyák, Domonkos Tikk, Szilveszter Kovács, Kok Wai Wong

Zs. Cs. Johanyák is with the Department of Information Technology, Kecskemét College, Kecskemét, H-6000 Hungary (e-mail: johanyak.csaba##AT##gamf.kefo.hu).

D. Tikk is with the Department of Telecommunications and Media Informatics, Budapest University of Technology and Economics, H-1117 Budapest, Hungary (e-mail:tikk##AT##tmit.bme.hu).

Sz. Kovács is with the Department of Information Technology, University of Miskolc, Miskolc, H-3515 Hungary (e-mail:szkovacs##AT##iit.uni-miskolc.hu).

K.W. Wong is with the School of Information Technology, Murdoch University, South St, Murdoch, Western Australia 6150 (e-mail:k.wong##AT##murdoch.edu.au)


Why Fuzzy Rule Interpolation is important?

< p class="Text" style="text-indent:0in">Fuzzy systems use fuzzy rule base to make inference. A fuzzy rule base is  fully covered (at level α ), if all input universes are covered by rules at level α. Such fuzzy rule bases are also called dense or complete rule bases. In practice, it means that for all the possible observations there exists at least one firing rule, whose antecedent part overlaps the input data at least partially at level α. If this condition is not satisfied, the rule base is considered sparse rule base, i.e. containing gaps. The classical fuzzy reasoning techniques like Zadeh’s, Mamdani’s, Larsen’s or even Sugeno’s cannot generate an acceptable output for such cases. Fuzzy rule based interpolation ( FRI) techniques were introduced to generate inference for sparse fuzzy rule base, thus extend the usage of sparse fuzzy rule base system. Basically, FRI techniques perform interpolative approximate reasoning by taking into consideration the existing fuzzy rules for cases where there is no fuzzy rules to fire.

The following are a list of papers on the background of Fuzzy Rule Interpolation.


< p>FRI Matlab toolbox Release 1.0

The details of the toolbox is available in the following paper: