Institute of Information Science, Academia Sinica

Graph Neural Networks (GNNs) have shown promising performance in many applications, yet remain extremely difficult to train over large-scale graph datasets. Existing weight pruning techniques can prune out the layer weights; however, they cannot fully address the high computation complexity of GNN inference, caused by large graph size and complicated node connections. In this paper, we propose an Iterative Gradient Rank Pruning (IGRP) algorithm to find graph lottery tickets (GLT) of GNNs where each GLT includes a pruned adjacency matrix and a sub-network. Our IGRP can avoid layer collapse and the winning ticket achieves Maximal critical compression. We evaluate the proposed method on smallscale (Cora and Citeseer), medium-scale (PubMed and Wiki-CS), and large-scale (Ogbn-ArXiv and Ogbn-Products) graph datasets. We demonstrate that both Single-shot and Multi-shot of IGRP outperform the state-of-the-art unified GNN sparsification (UGS) framework on node classification.

I am currently an assistant professor at the computer science and information engineering department from National Central University, Taoyuan City, Taiwan. I received my Ph.D. degree in 2024 from the Department of Computer Science and Software Engineering, Auburn University, Auburn, AL, USA. My B.S. degree is in mathematics and my M.S. degree is in computer science. I received them both from National Central University in 2013 and 2015. I joined the Institute for Information Industry, Taipei, Taiwan, where I served as an Engineer from 2015 to 2018. My research interest is Genetic Programming of Evolutionary Computation, Graph Neural Network of Machine Learning, and Searchable Encryption of Cybersecurity.