Efficient Multi-View Graph Clustering with Local and Global Structure ...
2023年10月27日 · To tackle the issue, we propose a novel anchor-based multi-view graph clustering framework termed Efficient Multi-View Graph Clustering with Local and Global Structure Preservation (EMVGC-LG). Specifically, a unified framework with a theoretical guarantee is designed to capture local and global information.
Two basic structures, i.e., global and local structures, both play crucial roles in numerous AMVGC studies. As shown in Figure 1 (right), the global structure, which captures the relationship be-tween samples and all anchors, usually has a dense correspondence matrix. Meanwhile, the matrix with local structure is always blocky
顶会论文看 图对比学习（GNN+CL）研究趋势 - 知乎
发表于 ICML 2021 的论文 Self-supervised Graph-level Representation Learning with Local and Global Structure 假设图的数据集中存在全局的、层次性的类别关联，并引入层次化的 Prototype 的概念，除了进行图表示之间的对比学习，还进行图表示与 prototype 表示之间的对比学习。
【论文笔记】Self-supervised Graph-level Representation Learning with Local ...
2022年7月3日 · 应当保留局部距离（local-instance structure），这样相似图可以进行嵌入，而不同的图则会被拆解； 应当反映全局语义（global-semantic structure），这样相似语义属性的图可以进行嵌入，有利于图分类或回归等各种下游任务。
Multi-view multi-manifold learning with local and global structure ...
2022年10月4日 · 1 Introduction. Multi-view learning obtains multiple views from multiple sources or different feature subsets. Exploring the consistency and complementarity of variety of different views, is more effective and promising than single-view …
Local and global structure preserving based feature selection
2012年7月15日 · In this paper, we propose two unsupervised FS methods called Local and Global Structure Preserving based Feature Selection (LGFS) and Extended LGFS (E-LGFS), respectively. LGFS firstly constructs a nearest neighborhood graph and a farthest neighborhood graph to describe the local and global structure of samples.
Self-supervised Graph-level Representation Learning with Local …
2021年6月8日 · In this paper, we propose a unified framework called Local-instance and Global-semantic Learning (GraphLoG) for self-supervised whole-graph representation learning. Specifically, besides preserving the local similarities, GraphLoG introduces the hierarchical prototypes to capture the global semantic clusters.
In this paper, we propose a unified framework called Local-instance and Global-semantic Learning (GraphLoG) for self-supervised whole-graph representation learning. Specifically, besides preserving the local similari-ties, GraphLoG introduces the hierarchical proto-types to capture the global semantic clusters.
Robust sparse and low-redundancy multi-label feature selection …
2023年2月1日 · To overcome these challenges, we advocate preserving both the dynamic local and global structure among labels to conduct feature selection for multi-label learning. Finally, a multi-label feature selection method named Sparse and Low-redundancy Multi-label feature selection with Dynamic local and global Structure preservation (SLMDS) is designed.
Unsupervised learning: Local and global structure preservation in ...
2023年10月1日 · PaCMAP is a dimensionality reduction technique that uses a stochastic gradient descent algorithm, boasts effective and reliable performance with preserving local and global structure alike, and performs significantly faster than other contemporaneous dimensionality reduction methods.
- 某些结果已被删除