Showing 1–2 of 2 results for author: Wong, Z H

Loss-aware Curriculum Learning for Heterogeneous Graph Neural Networks

Authors: Zhen Hao Wong, Hansi Yang, Xiaoyi Fu, Quanming Yao

Abstract: Heterogeneous Graph Neural Networks (HGNNs) are a class of deep learning models designed specifically for heterogeneous graphs, which are graphs that contain different types of nodes and edges. This paper investigates the application of curriculum learning techniques to improve the performance and robustness of Heterogeneous Graph Neural Networks (GNNs). To better classify the quality of the data,… ▽ More Heterogeneous Graph Neural Networks (HGNNs) are a class of deep learning models designed specifically for heterogeneous graphs, which are graphs that contain different types of nodes and edges. This paper investigates the application of curriculum learning techniques to improve the performance and robustness of Heterogeneous Graph Neural Networks (GNNs). To better classify the quality of the data, we design a loss-aware training schedule, named LTS that measures the quality of every nodes of the data and incorporate the training dataset into the model in a progressive manner that increases difficulty step by step. LTS can be seamlessly integrated into various frameworks, effectively reducing bias and variance, mitigating the impact of noisy data, and enhancing overall accuracy. Our findings demonstrate the efficacy of curriculum learning in enhancing HGNNs capabilities for analyzing complex graph-structured data. The code is public at https://github.com/LARS-research/CLGNN/. △ Less

Submitted 29 February, 2024; originally announced February 2024.

Ensemble Learning for Graph Neural Networks

Authors: Zhen Hao Wong, Ling Yue, Quanming Yao

Abstract: Graph Neural Networks (GNNs) have shown success in various fields for learning from graph-structured data. This paper investigates the application of ensemble learning techniques to improve the performance and robustness of Graph Neural Networks (GNNs). By training multiple GNN models with diverse initializations or architectures, we create an ensemble model named ELGNN that captures various aspec… ▽ More Graph Neural Networks (GNNs) have shown success in various fields for learning from graph-structured data. This paper investigates the application of ensemble learning techniques to improve the performance and robustness of Graph Neural Networks (GNNs). By training multiple GNN models with diverse initializations or architectures, we create an ensemble model named ELGNN that captures various aspects of the data and uses the Tree-Structured Parzen Estimator algorithm to determine the ensemble weights. Combining the predictions of these models enhances overall accuracy, reduces bias and variance, and mitigates the impact of noisy data. Our findings demonstrate the efficacy of ensemble learning in enhancing GNN capabilities for analyzing complex graph-structured data. The code is public at https://github.com/wongzhenhao/ELGNN. △ Less

Submitted 21 October, 2023; originally announced October 2023.