Survey on Deep Fuzzy Systems in Regression Applications: A View on Interpretability

Abstract

Deep learning (DL) has captured the attention of the community with an increasing number of recent papers in regression applications, including surveys and reviews. Despite the efficiency and good accuracy in systems with high-dimensional data, many DL methodologies have complex structures that are not readily transparent to human users. Accessing the interpretability of these models is an essential factor for addressing problems in sensitive areas such as cyber-security systems, medical, financial surveillance, and industrial processes. Fuzzy logic systems (FLS) are inherently interpretable models capable of using nonlinear representations for complex systems through linguistic terms with membership degrees mimicking human thought. This paper aims to investigate the state-of-the-art of existing deep fuzzy systems (DFS) for regression, i.e., methods that combine DL and FLS with the aim of achieving good accuracy and good interpretability. Within the concept of explainable artificial intelligence (XAI), it is essential to contemplate interpretability in the development of intelligent models and not only seek to promote explanations after learning (post hoc methods), which is currently well established in the literature. Therefore, this work presents DFS for regression applications as the leading point of discussion of this topic that is not sufficiently explored in the literature and thus deserves a comprehensive survey.