In this paper, we propose the first framework that enables solving graph learning tasks of all levels (node, edge and graph) and all types (generation, regression and classification) with one model. We first propose Latent Graph Diffusion (LGD), a generative model that can generate node, edge, and graph-level features of all categories simultaneously. We achieve this goal by embedding the graph structures and features into a latent space leveraging a powerful encoder which can also be decoded, then training a diffusion model in the latent space. LGD is also capable of conditional generation through a specifically designed cross-attention mechanism. Then we formulate prediction tasks including regression and classification as (conditional) generation, which enables our LGD to solve tasks of all levels and all types with provable guarantees. We verify the effectiveness of our framework with extensive experiments, where our models achieve state-of-the-art or highly competitive results across generation and regression tasks.

我们提出了第一个框架，允许使用一个模型解决所有级别（节点、边和图）和所有类型（生成、回归和分类）的图学习任务。我们首先提出了潜在图扩散（LGD），它是一个生成模型，可以同时生成所有类别的节点、边和图级特征。我们通过将图结构和特征嵌入到潜在空间中，并利用一个强大的编码器对其进行解码，然后在潜在空间中训练扩散模型来实现这个目标。LGD还能够通过一个特殊设计的交叉注意机制进行条件生成。然后，我们将包括回归和分类在内的预测任务制定为（条件）生成，使我们的LGD能够解决所有级别和所有类型的任务，并提供可证明的保证。通过广泛的实验证明了我们框架的有效性，其中我们的模型在生成和回归任务中达到了最先进或极具竞争力的结果。

潜在图扩散：一种在图上生成和预测的统一框架