An Introduction to Parameterized IFAM Models with Applications in Prediction
Peter Sussner, Rodolfo Miyasaki, Marcos Eduardo Valle · European Society for Fuzzy Logic and Technology Conference · 2009
Fuzzy associative memories (FAMs) and, in particu- lar, the class of implicative fuzzy associative memories (IFAMs) can be used to implement fuzzy rule-based systems. In this way, a vari- ety of applications can be dealt with. Since there are infinitely many IFAM models, we are confronted with the problem of selecting the best IFAM model for a given application. In this paper, we restrict ourselves to a subclass of the entire class of IFAMs, namely the sub- class of IFAMs that are associated with the Yager family of parame- terized t-norms. For simplicity, we speak of the class of Yager IFAMs. In this setting, we formulate the problem of choosing the best Yager IFAM for a given application as an optimization problem. Consid- ering two problems in time series prediction from the literature, we solve this optimization problem and compare the performance of the resulting Yager IFAM with the performances of other fuzzy, neural, neuro-fuzzy, and statistical techniques.