The integration of graph structures in diverse domains has recently garnered substantial attention, presenting a paradigm shift from classical euclidean representations. This new trend is driven by the advent of novel algorithms that can capture complex relationships through a class of neural architectures: the Graph Neural Networks (GNNs) [1], [2]. These networks are adept at handling data that can be effectively modeled as graphs, introducing a new representation learning paradigm. The significance of GNNs extends to several domains, including computer vision [3], [4], natural language processing [5], chemistry/biology [6], physics [7], traffic networks [8], and recommendation systems [9].

中文翻译：

---

客座社论基于图的方法和应用的新兴趋势和进展

不同领域中图结构的集成最近引起了广泛关注，呈现出经典欧几里得表示的范式转变。这一新趋势是由新颖算法的出现推动的，这些算法可以通过一类神经架构捕获复杂的关系：图神经网络 (GNN) [1]、[2]。这些网络擅长处理可以有效建模为图形的数据，引入新的表示学习范式。 GNN 的重要性扩展到多个领域，包括计算机视觉 [3]、[4]、自然语言处理 [5]、化学/生物学 [6]、物理学 [7]、交通网络 [8] 和推荐系统 [9] ]。