Hybridization of artificial neural network (ANN) and fuzzy logic (FL) has drawn the attention of researchers in various studies of scientific and engineering field due to the requirements of adaptive intelligent system for solving of real-world problems. Genetic algorithm (GA) has been frequently used to optimize the problem solutions. ANN imitate the work principles of human brain, and realize the learning via using the samples in training process. FL converts the linguistic expressions to rules in a rule base via using given rules and membership functions. When ANN works in conjunction with FL to fill lacks, high performance systems are obtained. The learning ability can be added to FL-based systems via ANN usage. In neuro-fuzzy systems (NFSs), the ability of flexibility, speed and adaptivity can be fused to FL component through ANN component. In our study, 51 studies in the literature about NFSs are systematically reviewed. These studies are based on the hybridization of ANN and FL components. As can be seen from the survey, the approaches based on the adaptive neural fuzzy inference system (ANFIS) are much more used than other neuro-fuzzy systems’studies in the literature. We made a conclusion over example works in the literature.
Özet. Yapay Sinir Ağları (Artificial Neural Network, YSA) ve Bulanık Mantık (Fuzzy Logic, BM) melezleştirmesi gerçek dünya problemlerinin çözümünde uyarlanabilir zeki sistemlere olan ihtiyaç nedeniyle çeşitli bilimsel ve mühendislik alanındaki çalışmalarda araştırmacıların ilgisini çekmektedir. Problemlerin çözümünde sıklıkla eniyileme için Genetik Algoritma (Genetic Algorithm, GA) kullanılmaktadır. YSA, insan beyninin çalışma prensibini taklit ederek, eğitim sürecindeki örneklerin kullanımı sayesinde öğrenimini gerçekleştirir. BM, sözel ifadeleri verilen kurallar ve üyelik fonksiyonları kullanarak kural tabanındaki kurallara çevirmektedir. YSA ve BM birbirlerinin eksikliklerini giderdiklerinde başarımı daha yüksek sistemler elde edilmektedir. Bulanık sistemlere sinir ağı ile öğrenme yeteneği kazandırılabilmektedir. Sinirsel bulanık sistemlerde (SBS), BM bileşenine esneklik, hız ve uyarlanabilirlik gibi özellikler YSA bileşeni sayesinde kaynaştırılmaya çalışılmaktadır. Bu çalışmada, YSA ve BM bileşenlerinin melezlenmesiyle elde edilmiş literatürdeki SBS’lerle ilgili 51 adet çalışma sistematik olarak incelenmiştir. Yapılan literatür incelenmesinde Uyarlanabilir Sinirsel Bulanık Çıkarım Sistemini (Adaptive Neural Fuzzy Inference System, ANFIS) temel alan yaklaşımların diğer SBS’lere göre daha fazla sayıda çalışmada kullanıldığı görülmektedir. Literatürdeki örnek çalışmalar üzerinden değerlendirme yapılmıştır.
Abstract. Hybridization of artificial neural network (ANN) and fuzzy logic (FL) has drawn the attention of researchers in various studies of scientific and engineering field due to the requirements of adaptive intelligent system for solving of real-world problems. Genetic algorithm (GA) has been frequently used to optimize the problem solutions. ANN imitate the work principles of human brain, and realize the learning via using the samples in training process. FL converts the linguistic expressions to rules in a rule base via using given rules and membership functions.
When ANN works in conjunction with FL to fill lacks, high performance systems are obtained. The learning ability can be added to FL-based systems via ANN usage. In neuro-fuzzy systems (NFSs), the ability of flexibility, speed and adaptivity can be fused to FL component through ANN component. In our study, 51 studies in the literature about NFSs are systematically reviewed. These studies are based on the hybridization of ANN and FL components. As can be seen from the survey, the approaches based on the adaptive neural fuzzy inference system (ANFIS) are much more used than other neuro-fuzzy systems’studies in the literature. We made a conclusion over example works in the literature.
Primary Language | Turkish |
---|---|
Journal Section | Editorial |
Authors | |
Publication Date | May 8, 2014 |
Published in Issue | Year 2014 Volume: 35 Issue: 2 |