<p>Deep learning technologies have achieved significant advancements across various domains and have been widely integrated into everyday life. However, traditional deep learning models rely on vast datasets with precise annotations, which often require professional knowledge and vast time. On the one hand, acquiring relevant data, particularly in specialized fields such as healthcare, is a significant challenge. On the other hand, even when we get enough data, the precise annotation of vast datasets remains a significant challenge. Few-shot learning (FSL) techniques are crucial. Consequently, the paper proposes an Implicit Symbolic Rule Generator (ISRG) module that integrates deep learning with symbolic knowledge to enhance the model’s relational reasoning performance in FSL. The module transforms image features into implicit logical predicates, then constructs implicit logical rules using symbolic connection. By selecting an appropriate set of rules as input modalities, the module enhances the training efficiency and accelerates the speed of loss convergence of the model. Experimental results from four few-shot datasets demonstrate that the module not only significantly improves performance but also accelerates the convergence rate of loss. The module is likely to exhibit superior performance when applied to few-shot datasets with complex image features.</p>