Debiased Graph Neural Networks With Agnostic Label Selection Bias
- 1 February 2022
- journal article
- research article
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Neural Networks and Learning Systems
- Vol. 35 (4) , 4411-4422
Abstract
Most existing graph neural networks (GNNs) are proposed without considering the selection bias in data, i.e., the inconsistent distribution between the training set with the test set. In reality, the test data are not even available during the training process, making selection bias agnostic. Training GNNs with biased selected nodes leads to significant parameter estimation bias and greatly impacts the generalization ability on test nodes. In this article, we first present an experimental investigation, which clearly shows that the selection bias drastically hinders the generalization ability of GNNs, and theoretically proves that the selection bias will cause the biased estimation on GNN parameters. Then to remove the bias in GNN estimation, we propose a novel debiased GNNs (DGNN) with a differentiated decorrelation regularizer. The differentiated decorrelation regularizer estimates a sample weight for each labeled node such that the spurious correlation of learned embeddings could be eliminated. We analyze the regularizer in causal view and it motivates us to differentiate the weights of the variables based on their contribution to the confounding bias. Then, these sample weights are used for reweighting GNNs to eliminate the estimation bias, and thus, help to improve the stability of prediction on unknown test nodes. Comprehensive experiments are conducted on several challenging graph datasets with two kinds of label selection biases. The results well verify that our proposed model outperforms the state-of-the-art methods and DGNN is a flexible framework to enhance existing GNNs.Keywords
All Related Versions
Funding Information
- National Natural Science Foundation of China (U20B2045, 62192784, 62172052, 61772082, 62002029)
- Fundamental Research Funds for the Central Universities (2021RC28)
- Beijing University of Posts and Telecommunications Excellent Ph.D. Students Foundation (CX2021311)
- China Scholarship Council
- National Natural Science Foundation of China (62006207)
- Young Elite Scientists Sponsorship Program by CAST
- Zhejiang Province Natural Science Foundation (LQ21F020020)
This publication has 25 references indexed in Scilit:
- Data-Driven Variable Decomposition for Treatment Effect EstimationIEEE Transactions on Knowledge and Data Engineering, 2020
- One2Multi Graph Autoencoder for Multi-view Graph Clustering2020
- Causal InferenceEngineering, 2020
- Metapath-guided Heterogeneous Graph Neural Network for Intent Recommendation2019
- Stable Prediction across Unknown Environments2018
- Decorrelation of Neutral Vector Variables: Theory and ApplicationsIEEE Transactions on Neural Networks and Learning Systems, 2016
- Evaluation of the Effect of a Continuous Treatment: A Machine Learning Approach with an Application to Treatment for Traumatic Brain InjuryHealth Economics, 2015
- Entropy Balancing for Causal Effects: A Multivariate Reweighting Method to Produce Balanced Samples in Observational StudiesPolitical Analysis, 2012
- Absolute and relative Weyl theorems for generalized eigenvalue problemsLinear Algebra and its Applications, 2009
- The Graph Neural Network ModelIEEE Transactions on Neural Networks, 2008