New Framework Boosts Interpretable Fuzzy Learning with Meta-Learning

Achieving interpretable representation learning in neural networks remains a significant challenge, especially when models need to explain their reasoning. This paper introduces iFuzz-Meta, an interpretable fuzzy rule-based learning framework designed to maintain human-understandable reasoning structures within modern neural architectures. Each fuzzy rule in iFuzz-Meta corresponds to a semantic and spatial prototype, allowing for transparent inference. The framework employs meta-learning to analyze how these interpretable rules adapt across different tasks and domains, linking algorithmic adaptation to cognitive representation. A knowledge-guided regularization mechanism further integrates theoretical priors as soft inductive biases with data-driven learning, ensuring adaptation follows meaningful trajectories. Evaluations demonstrate that iFuzz-Meta provides both interpretable reasoning and stable cross-domain generalization, offering a promising path toward explainable and knowledge-aware fuzzy systems.