ARDiff: Anisotropic Residual Diffusion for Heterogeneous Graph Learning

Abstract

Learning representations on graphs is foundational for many downstream tasks, and its synergy with diffusion models has emerged as a promising direction. However, diffusion-based methods for heterogeneous graphs remain underexplored, confronting two principal challenges: (1) The presence of noise and structural heterogeneity in graphs makes it challenging to accurately capture semantic transitions among diverse relation types. (2) The isotropic Gaussian noise used in forward diffusion fails to reflect graphs' inherent semantics and structural anisotropy. To address these, we propose ARDiff, a novel framework that integrates residual diffusion with anisotropic noise for heterogeneous graph learning. Specifically, we propose a semantic residual diffusion mechanism that progressively refines node embeddings by orchestrating transitions from low-semantic (high-noise) to high-semantic (low-noise) relational contexts, thus enabling step-wise distillation of task-relevant information. In addition, to address the limitations of conventional diffusion, we introduce an anisotropic diffusion strategy: in the forward process, noise injection is oriented by structural and semantic priors; in the denoising step, a conditional diffusion mechanism is guided by a random walk encoding, enhancing both topological consistency and semantic alignment. Extensive evaluation on heterogeneous graph datasets demonstrates that ARDiff significantly surpasses current leading methods in link prediction and node classification, setting a new paradigm and benchmark in heterogeneous graph representation learning.

Environment

• PyTorch 2.1.2
• Python 3.10(ubuntu22.04)
• CUDA 11.8
• dgl 2.1.0+cu118

Code Structures

The folder ARDiff-LP presents the code and datasets for link prediction, while ARDiff_NC for node classification.

.
├──ARDiff-LP
 ├── DataHandler.py
 ├── main.py
 ├── param.py
 ├── Utils                    
 │   ├── TimeLogger.py            
 │   ├── Utils.py                             
 ├── Model.py
├──ARDiff-NC
 ├──DataHandler.py
 ├── main.py
 ├── param.py
 ├── Utils                    
 │   ├── TimeLogger.py            
 │   ├── Utils.py
 ├── Model.py
└── README

Datasets

We evaluate ARDiff on both the link prediction and node classification tasks. For link prediction, we utilize three publicly available datasets collected from real-world commercial platforms: Tmall, Retailrocket, and IJCAI. For the node classification task, we use two public datasets, DBLP and AMiner, which focus on publications and academic social ties. Below are the detailed descriptions for the experimental datasets.

Dataset	User #	Item #	Link #	Interaction Types
Retail Rocket	2174	30113	97,381	View, Cart, Transaction
IJCAI	17435	35920	799,368	View, Favorite, Cart, Purchase
Tmall	31882	31232	1,451,29	View, Favorite, Cart, Purchase

	Node	Metapath		Node	Metapath
DBLP	Author:4057	APA	AMiner	paper:6564	PAP
	Paper:14328	APCPA		author:13329	PRP
	Conference:20	APTPA		Reference:35890	POS
	Term:7723