Graph attention networks have demonstrated competitive performance in node classification tasks. Nonetheless, recent studies have highlighted their limitation in generating static attention scores. Static attention coefficients imply that attention scores for node pairs remain fixed regardless of variational query keys or structural information. Consequently, researchers have explored diverse strategies to devise a dynamic attention mechanism suitable for graph data. In this vein, we introduce a dynamic attention-based graph attention network (GAT) grounded in structural information learning and engaged in enhancing its efficacy.