Fine-tuning Graph Neural Networks by Preserving Graph Generative Patterns

21 Dec 2023  ·  Yifei Sun, Qi Zhu, Yang Yang, Chunping Wang, Tianyu Fan, Jiajun Zhu, Lei Chen ·

Recently, the paradigm of pre-training and fine-tuning graph neural networks has been intensively studied and applied in a wide range of graph mining tasks. Its success is generally attributed to the structural consistency between pre-training and downstream datasets, which, however, does not hold in many real-world scenarios. Existing works have shown that the structural divergence between pre-training and downstream graphs significantly limits the transferability when using the vanilla fine-tuning strategy. This divergence leads to model overfitting on pre-training graphs and causes difficulties in capturing the structural properties of the downstream graphs. In this paper, we identify the fundamental cause of structural divergence as the discrepancy of generative patterns between the pre-training and downstream graphs. Furthermore, we propose G-Tuning to preserve the generative patterns of downstream graphs. Given a downstream graph G, the core idea is to tune the pre-trained GNN so that it can reconstruct the generative patterns of G, the graphon W. However, the exact reconstruction of a graphon is known to be computationally expensive. To overcome this challenge, we provide a theoretical analysis that establishes the existence of a set of alternative graphons called graphon bases for any given graphon. By utilizing a linear combination of these graphon bases, we can efficiently approximate W. This theoretical finding forms the basis of our proposed model, as it enables effective learning of the graphon bases and their associated coefficients. Compared with existing algorithms, G-Tuning demonstrates an average improvement of 0.5% and 2.6% on in-domain and out-of-domain transfer learning experiments, respectively.

PDF Abstract

Results from the Paper


 Ranked #1 on Graph Classification on IMDb-M (Accuracy (10-fold) metric)

     Get a GitHub badge
Task Dataset Model Metric Name Metric Value Global Rank Benchmark
Graph Classification BACE G-Tuning ROC-AUC 84.79 # 1
Graph Classification BBBP G-Tuning ROC-AUC 72.59 # 1
Graph Classification clintox G-Tuning ROC-AUC 74.64 # 1
Graph Classification ENZYMES G-Tuning Accuracy (10-fold) 26.70 # 2
Graph Classification HIV G-Tuning ROC-AUC 77.33 # 3
Graph Classification IMDb-B G-Tuning Accuracy (10-fold) 74.30 # 3
Graph Classification IMDb-M G-Tuning Accuracy (10-fold) 51.80 # 1
Graph Classification MSRC-21 (per-class) G-Tuning Accuracy (10 fold) 11.01 # 1
Graph Classification MUTAG G-Tuning Accuracy (10 fold) 86.14 # 1
Graph Classification MUV G-Tuning ROC-AUC 75.84 # 2
Graph Classification PROTEINS G-Tuning Accuracy (10 fold) 72.05 # 2
Graph Classification REDDIT-12K G-Tuning Accuracy (10 fold) 42.80 # 1
Graph Classification SIDER G-Tuning ROC-AUC 61.40 # 2
Graph Classification Tox21 G-Tuning ROC-AUC 75.80 # 2
Graph Classification ToxCast G-Tuning ROC-AUC 64.25 # 2

Methods