Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

Universal Prompt Tuning for Graph Neural Networks

About

In recent years, prompt tuning has sparked a research surge in adapting pre-trained models. Unlike the unified pre-training strategy employed in the language field, the graph field exhibits diverse pre-training strategies, posing challenges in designing appropriate prompt-based tuning methods for graph neural networks. While some pioneering work has devised specialized prompting functions for models that employ edge prediction as their pre-training tasks, these methods are limited to specific pre-trained GNN models and lack broader applicability. In this paper, we introduce a universal prompt-based tuning method called Graph Prompt Feature (GPF) for pre-trained GNN models under any pre-training strategy. GPF operates on the input graph's feature space and can theoretically achieve an equivalent effect to any form of prompting function. Consequently, we no longer need to illustrate the prompting function corresponding to each pre-training strategy explicitly. Instead, we employ GPF to obtain the prompted graph for the downstream task in an adaptive manner. We provide rigorous derivations to demonstrate the universality of GPF and make guarantee of its effectiveness. The experimental results under various pre-training strategies indicate that our method performs better than fine-tuning, with an average improvement of about 1.4% in full-shot scenarios and about 3.2% in few-shot scenarios. Moreover, our method significantly outperforms existing specialized prompt-based tuning methods when applied to models utilizing the pre-training strategy they specialize in. These numerous advantages position our method as a compelling alternative to fine-tuning for downstream adaptations.

Taoran Fang, Yunchao Zhang, Yang Yang, Chunping Wang, Lei Chen• 2022

Related benchmarks

TaskDatasetResultRank
Node ClassificationCora
Accuracy45.16
1215
Node ClassificationCora (test)
Mean Accuracy43.16
861
Node ClassificationCiteseer
Accuracy24.88
393
Node ClassificationPhoto
Mean Accuracy55.1
343
Node Classificationogbn-products (test)
Test Accuracy13.43
137
Graph ClassificationIMDB-B (test)
Accuracy73.4
134
Graph ClassificationCIFAR10
Accuracy62.3
110
Graph property predictionTox21
ROC-AUC0.8463
109
Graph ClassificationREDDIT BINARY
Accuracy72.5
107
Graph ClassificationHIV
ROC-AUC0.7912
104
Showing 10 of 64 rows

Other info

Code

Follow for update