在本文中,作者研究了对 GAE 发展产生负面影响的关键问题——重建目标、训练鲁棒性、误差指标,然后提出了GraphMAE(masked graph autoencoder),这缓解了生成自监督图预训练的这些问题。作者建议不是重建图结构,而是使用掩蔽策略专注于特征重建,损失函数缩放余弦误差(Scaled cosine error),有利于 GraphMAE 的稳健训练。 模...
graph masked autoencodergraph contrastive learningSpatial transcriptomics (ST) technology provides gene expression profiles with spatial context, offering critical insights into cellular interactions and tissue architecture. A core task in ST is spatial domain identification, which involves detecting coherent ...
Masked feature reconstruction 与大多数GAE在结构重建方面的研究方向不同,GraphMAE只关注于掩蔽特征的重建,其有效性已在CV和NLP 中得到广泛验证。实证研究表明,通过适当的误差设计,掩蔽特征重建可以极大地有利于GAE。 Scaled cosine error GraphMAE不使用MSE,而是使用余弦误差,当特征向量的大小发生变化时,这是有益的(图...
To solve this problem, we propose a new framework called Graph Clustering with Masked Autoencoders (GCMA). It employs our designed fusion autoencoder based on the graph masking method for the fusion coding of graph. It introduces our improved density-based clustering algorithm as a second ...
论文标题:Graph Masked Autoencoders with Transformers论文作者:Sixiao Zhang, Hongxu Chen, Haoran Yang, Xiangguo Sun, Philip S. Yu, Guandong Xu论文来源:2022, ArXiv论文地址:download 论文代码:download 1 Introduction提出目的:深层Tramsformer 的困难; 指数级的内存消耗;...
Graph masked autoencoder for sequential recommendation. SIGIR, 2023. 概 图+ MAE. 符号说明 U,VU,V, users, items; Su=(su1,su2,⋯,sulu)Su=(s1u,s2u,⋯,sluu), 某个序列; G=(V,E)G=(V,E), global item transition graph, 这里 E={(sut,sut′):u∈U,|t−t′|≤h,1≤t,t′...
Masked Graph Autoencoders with Contrastive Augmentation for Spatially Resolved Transcriptomics Data - wenwenmin/STMGAC
Deep Graph Clustering with Unknown Cluster Number YearTitleVenuePaperCode 2024Masked AutoEncoder for Graph Clustering without Pre-defined Cluster Number k (GCMA)arXivLink- 2023Reinforcement Graph Clustering with Unknown Cluster Number (RGC)ACM MMLinkLink ...
GraphMAE: Self-Supervised Masked Graph Autoencoders 论文摘要 近年来,人们对自监督学习(SSL)进行了广泛的研究。特别是,生成性SSL已经在自然语言处理和其他领域取得了成功,例如BERT和GPT的广泛采用。尽管如此,对比学习(严重依赖于结构数据增强和复杂的训练策略)一直是图SSL的主要方法,而生成性SSL在图上的进展,特别是...
Node clustering Ablation experiments 来自 美国圣母大学 论文代码github.com/meettyj/HGMAE Introduction Generative self-supervised learning (SSL),尤其是masked autoencoders在图数据挖掘中展现出巨大潜力。本文研究了异构图上的生成式SSL问题,并提出了一种异构图掩码自编码器模型——HGMAE。针对异构图,提出以下训...