GitHub is where people build software. More than 150 million people use GitHub to discover, fork, and contribute to over 420 million projects.
本文认为目前的一些图自编码器方法(Graph Autoencoder, GAE)面临以下四个问题,虽然一些问题解决了其中的一两个,但是没有一个方法将这个四个问题作为一个整体来对待: ①结构信息被过分强调了,大多数GAE方法以连接重构(link reconstruction)来作为目标,因此之前的GAE方法在连接预测和节点聚类任务上很擅长,但是在图和节点...
论文链接:One2Multi Graph Autoencoder for Multi-view Graph Clustering 论文源码:https://github.com/songzuolong/WWW2020-O2MAC 提出背景 前人研究multi-view的方法可以分为两类: 基于图分析方法,最大化不同view之间的某种相互协议,然后将一个图划分为多个组 基于图嵌入方法,从multi-view中学习紧凑的节点表示 这...
Masked Graph Autoencoders with Contrastive Augmentation for Spatially Resolved Transcriptomics Data Introduction In this paper, we propose a Contrastively Augmented Masked Graph Autoencoder (STMGAC) to learning low-dimensional latent representations for SRT data analysis. First, we use a masked graph aut...
The source code of GDAE can be found in https://github.com/luliu-fighting/Graph-Dynamic-Autoencoder .doi:10.1016/j.ces.2022.117637Liu, LuZhao, HaitaoHu, ZhengweiChemical Engineering Science
论文标题:GraphMAE: Self-Supervised Masked Graph Autoencoders 论文链接:https://arxiv.org/abs/2205.10803 论文来源:KDD 2022 代码链接:https://github.com/THUDM/GraphMAE 一、概述 自监督学习(Self-supervised learning, SSL)通常分为生成式学习和对比学习。目前图领域的自监督学习以对比学习的方法为主,尤其是...
Autoencoders, a type of unsupervised model, are capable of learning effective latent representations of data without supervision, only requiring the decoder to be able to reconstruct the original data-point from its latent representation obtained through the encoder. When dealing with structured data,...
实现的模型代码可在https://github.com/HKUDS/MAERec. 1.简介 顺序推荐旨在学习用户兴趣的有效表示,并建议不同用户可能感兴趣的未来项目。这项任务引起了例如电子商务和视频流网站相当大的关注。因为用户偏好在现实生活中是时间演变的,为了捕捉项目之间的高阶转换关系,大量的研究工作致力于提出各种图神经网络,通过在...
inconsistent spatial clusters across mouse samples: cells from the same cluster (blue) correspond to different regions in 13-months control (left) and 13-months AD (right) mice. The latent dimension of this model is 1024.dClustering of the cells in the latent space by our autoencoder model ...
It improved the graph autoencoder to preserving global topological structure among cells. We further extended the scGAE for visualization, trajectory inference, and clustering. Analyses of simulated data and empirical data showed that scGAE outperformed the other competitive methods....