Review of Deep Learning Revolution on Neuro-Fuzzy Systems

Hybrid intelligent systems, also known as neuro-fuzzy systems, are a fusion of fuzzy logic with neural networks. They hold immense potential in tackling complex tasks and are often prone to error and ambiguity, sparking curiosity in their capabilities. This review paper delves into the foundations, architectures, applications, challenges, and future research possibilities published between 2018 and 2023. The fundamentals of neural networks, fuzzy logic, and their integration are explored, inviting the reader to delve deeper into these concepts. This work also discusses important neuro-fuzzy system architectures, such as Takagi–Sugeno models and ANFIS, highlighting their significance. It also assesses the development of Deep Neuro-Fuzzy Systems (DNFSs), integrating deep learning techniques with neuro-fuzzy principles to provide a workable remedy for the drawbacks of conventional methods. Neuro-fuzzy systems can automatically extract hierarchical features and achieve end-to-end optimization of membership functions, rules, and parameters through deep learning, demonstrating their potential. The architecture of DNFSs is assessed using either Deep Neural Networks (DNNs) or Convolutional Neural Networks (CNNs), showcasing their versatility. Around 70–80 papers were initially reviewed for the paper, indicating the extensive research in this field. This review paper highlights these limitations and the potential for further study, stimulating the reader’s interest in future developments. Finally, the authors investigate neuro-fuzzy and DNFSs in applications that require uncertainty modeling or data-driven reasoning, emphasizing their real-world applications. Patterns and neuro-fuzzy systems promise to produce brilliant, interpretable models, sparking the reader’s imagination about their potential.