Relational Modular Fuzzy Systems

摘要: This chapter presents the fuzzy relational model. In such model we define all possible connections between input and output linguistic terms [6, 12]. An advantage of this approach is great flexibility system. Input are fully interconnected. Moreover, can be modeled by changing elements relation matrix. The matrix regarded as a set similar to rule weights in classic systems [8, 11]. Relational used successfully e.g. control [4] classification tasks [1, 21, 23]. chapter, neuro-fuzzy [17, 20] will used. Such neural network like structures allow use more scenarios than case ordinary structures. For example, values advance then fine tune mapping using gradient learning. Gradient learning an important comparing original systems. Furthermore, AdaBoost ensembles classifiers. A serious drawback system boosting that contain separate bases which cannot directly merged. As separate, treat rules coming from different same (single) problem addressed novel design constituting ensemble, resulting normalization individual during There were some attempts combine models, [13] or rough-fuzzy models [9] but none them solved multiple ensemble.