10 February 2023 in IEEE Transactions on Fuzzy Systems
Explainability of deep neural networks has been receiving increasing attention with regard to auditability and trustworthiness purposes. Of the various post-hoc explainability approaches, rule extraction methods assist to understand the logic that underpins their functioning. Whereas the rule-based solutions are directly managed and understood by practitioners, the use of intervals or crisp values in the antecedents that rely on numerical values might not be intuitive enough. In this case, the benefits of a linguistic representation based on fuzzy sets/rules are straightforward, as these semantically meaningful components ease the model understanding. This paper proposes fuzzy rule-based explainer systems for deep neural networks. The algorithm learns a compact yet accurate set of fuzzy rules based on features' importance (i.e., attribution values) distilled from the trained networks. These systems can be used for both local and global explainability purposes. The evaluation results of different applications revealed that the fuzzy explainers maintained the fidelity and accuracy of the original deep neural networks while implying lower complexity and better comprehensibility.
1 September 2019 in Information Sciences
1 December 2020 in Expert Systems with Applications
3 February 2026 in Multimedia Tools and Applications
8 January 2021 in IEEE Transactions on Fuzzy Systems
In current Data Science applications, the course of action has derived to adapt the system behavior for the human cognition, resulting in the emerging area of explainable artificial intelligence. Among different classification paradigms, those based on fuzzy rules are suitable solutions to stress the interpretability of the global systems. However, in case of addressing Big Data analytics, they may comprise an excessive number of rules and/or linguistic labels that not only may cause losing the system performance but also may affect the system semantic as well as the system interpretability. In this article, we propose IFC-BD, an interpretable fuzzy classifier for Big Data, aiming at boosting the horizons of explainability by learning a compact yet accurate fuzzy model. IFC-BD is developed in a cell-based distributed framework through the three working stages of initial rule learning, rule generalization, and heuristic rule selection. This whole procedure allows reaching from a high number of specific rules to less number of more general and confident rules. Additionally, in order to resolve possible rules conflict, a new estimated rule weight is proposed specifically for big data problems. IFC-BD was evaluated in comparison to the state-of-the-art approaches of the fuzzy classification paradigm, considering interpretability, accuracy, and running time. The findings of the experiments revealed that the proposed algorithm was able to improve the explainability of fuzzy rule-based classifiers as well as their predictive performance.
1 June 2019 in Applied Soft Computing