Graph-based neural network models with multiple self-supervised auxiliary tasks
Graph-based neural network models with multiple self-supervised auxiliary tasks
复制标题
DOI：
10.1016/j.patrec.2021.04.021
复制
发表时间：
2021-05-24
影响因子：
5.1
通讯作者：
Rozza, Alessandro
中科院分区：
文献类型：
--
作者：
Manessi, Franco;Rozza, Alessandro
自监督学习目前获得了很多关注，因为它允许神经网络从大量未标记的数据中学习稳健表示。此外，多任务学习可以通过在相关任务上同时训练网络来进一步改进表征学习，从而显著提高性能。在本文中，我们提出了三个新的自监督辅助任务，以多任务的方式训练基于图的神经网络模型。由于图卷积网络是捕获结构化数据点之间关系的最有前途的方法之一，因此我们将其用作构建块，以在标准半监督图分类任务上获得竞争性结果。(c) 2021 Elsevier B.V.版权所有
Self-supervised learning is currently gaining a lot of attention, as it allows neural networks to learn ro-bust representations from large quantities of unlabeled data. Additionally, multi-task learning can further improve representation learning by training networks simultaneously on related tasks, leading to signifi-cant performance improvements. In this paper, we propose three novel self-supervised auxiliary tasks to train graph-based neural network models in a multi-task fashion. Since Graph Convolutional Networks are among the most promising approaches for capturing relationships among structured data points, we use them as a building block to achieve competitive results on standard semi-supervised graph classifi-cation tasks. (c) 2021 Elsevier B.V. All rights reserved.