11 Jun 2023·Hongyi Ling,Zhimeng Jiang,Meng Liu,Shuiwang Ji,Na Zou· We study graph data augmentation by mixup, which has been used successfully on images. A key operation of mixup is to compute a convex combination of a pair of inputs. This operation is straightforward for grid-like data...
Intrusion-Free Graph MixupHongyu GuoYongyi Mao
We present a simple and yet effective interpolation-based regularization technique to improve the generalization of Graph Neural Networks (GNNs). We leverage the recent advances in Mixup regularizer for vision and text, where random sample pairs and their labels are interpolated to create synthetic ...
iGraphMix: Input Graph Mixup Method for Node Classificationopenreview.net/forum?id=a2ljjXeDcE Introduction Mixup 是一种有效的数据增强方法,通过从训练集中随机选择两个样本和线性插值样本和相应的标签来增强数据。有助于防止神经网络的过拟合问题,特别是由于在训练中缺乏标记数据而引起的问题。 由于图神经网络...
图1:(左)对于图像分类,现有 Mixup 通过对图像像素和标签进行插值来生成合成图像。 (中)对于节点分类,要混合一对节点 A(红色)和 B(蓝色),我们需要混合它们的感受野子图。 (右)对于图分类,我们需要混合一对图的节点和图拓扑。 虽然Mixup在增强图像数据方面是有效的,但为图学习设计Mixup方法是具有挑战性的。挑战的...
1. 首先,本文提出了 G-Mixup 来扩充用于图分类的训练图。由于直接混合图是难以处理的,因此 G-Mixup 将不同类别的图的图元混合以生成合成图。 2. 其次,本文从理论上证明合成图将是原始图的混合,其中源图的关键拓扑(即判别主题)将被混合。 3. 最后,本文证明了所提出的 G-Mixup 在各种图神经网络和数据集上的...
We study graph data augmentation by mixup, which has been used successfully on images. A key operation of mixup is to compute a convex combination of a pair of inputs. This operation is straightforward for grid-like data, such as images, but challenging for graph data. The key difficulty ...
This paper presents GraphMixup, a novel mixup-based framework for improving class-imbalanced node classification on graphs. However, directly performing mixup in the input space or embedding space may produce out-of-domain samples due to the extreme sparsity of minority classes; hence we construct ...
在这项研究中,作者提出了一种新的图数据增广方法:G-Mixup,实验表明,G-Mixup 能够提高图神经网络的泛化性和鲁棒性。 2.1.1 动机 当前流行的数据增广方法 Mixup 通过在两个随机样本之间插入特征和标签,在提高神经网络的泛化性和鲁棒性方面显示出...
Graph transplant: Node saliency-guided graph mixup with local structure preservation (AAAI 2022) main.py : Graph Transplant main_aug.py : dropN, permE, maskN, subG, M-evolve @inproceedings{park2022graph, title={Graph transplant: Node saliency-guided graph mixup with local structure preservation...