Hybrid learning-based neuro-fuzzy inference system: a new approach for system modeling

Abstract

In this article, a hybrid learning neuro-fuzzy inference system (HLNFIS) with a new inference mechanism is proposed for system modeling. In the HLNFIS, the incoming signal is fuzzified by the proposed improved Gaussian membership function (IGMF), which is derived from two standard Gaussian functions. With the premise construction with IGMFs, the system inference ability can be upgraded. The fuzzy inference processor, which involves both numerical and linguistic reasoning, is introduced in rule base construction. For effective parameter learning, the hybrid algorithm of random optimization (RO) and least square estimation (LSE) is exploited, where the premise and the consequence parameters of are updated by RO and LSE, respectively. To validate the feasibility and the potential of the proposed approach, three examples of system modeling are conducted. Through experimental results and comparisons the proposed HLNFIS shows excellent performance for complex modeling.

Fuzzy and SVM Based Classification Model to Classify Spectral Objects in Sloan Digital Sky

“…At the close of the past decade, a Type-2 Fuzzy Logic System (T2FLS) was suggested that employs nearly two fuzzy Membership Functions (MFs), which increases the capability to handle language uncertainty representation [33] [34]. T2FS and T2FLS are used in the following applications: (i) control and system modelling [35] [36], (ii) robots and motion control [37] [38], and (iii) image processing [39] [40]. When the systems are subjected to various uncertainties, T2FLS outperforms standard Type-1 fuzzy systems.…”

Development of a rule selection mechanism by using neuro-fuzzy methodology for structural vibration suppression

“…Zaheeruddin and Garima [19] proposed a neuro-fuzzy approach for predicting work efficiency in noisy environments. Cheng [20] applied a hybrid neuro-fuzzy inference system, and Banakar and Azeem [21] also presented a method for the identification of nonlinear dynamical plants using a neuro-fuzzy model. To attain the interpretability of neuro-fuzzy systems, Paiva and Dourado [22] proposed a method that imposes some constraints on the tuning of the parameters and merges membership functions.…”