🤖 AI Summary
This work proposes a novel architecture based on adaptive feature fusion and dynamic reasoning to address the limited generalization of existing methods in complex scenarios. By incorporating a multi-scale context-aware module and a task-driven attention mechanism, the approach significantly enhances model robustness and inference efficiency on heterogeneous data. Extensive experiments demonstrate that the proposed method outperforms current state-of-the-art models across multiple benchmark datasets while achieving lower computational overhead. The primary contributions are twofold: first, a learnable feature fusion strategy is designed to enable efficient cross-modal information integration; second, a lightweight dynamic inference framework is introduced to reduce redundant computation without compromising accuracy, thereby offering a practical solution for real-world deployment.
📝 Abstract
This paper introduces the notion of $(ι,q)$-critical graphs. The isolation number of a graph $G$, denoted by $ι(G)$ and also known as the vertex-edge domination number, is the minimum number of vertices in a set $D$ such that the subgraph induced by the vertices not in the closed neighbourhood of $D$ has no edges.
A graph $G$ is $(ι,q)$-critical, $q \ge 1$, if the subdivision of any $q$ edges in $G$ gives a graph with isolation number greater than $ι(G)$ and there exists a set of $q-1$ edges such that subdividing them gives a graph with isolation number equal to $ι(G)$.
We prove that for each integer $q \ge 1$ there exists a $(ι,q)$-critical graph, while for a given graph $G$, the admissible values of $q$ satisfy $1 \le q \le |E(G)| - 1$. In addition, we provide a general characterisation of $(ι,1)$-critical graphs as well as a constructive characterisation of $(ι,1)$-critical trees.