Hypergraph Attention Networks

Chen, Chaofan; Cheng, Zelei; Li, Zuotian; Wang, Manyi

doi:10.1109/trustcom50675.2020.00215

Cited by 15 publications

(19 citation statements)

References 20 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Message propagation on the hypergraph is more complex than the graph, involving two stages: attentive hyperedge aggregation and attentive vertex aggregation [41]. In the attentive hyperedge aggregation, the connected vertexes in the same hyperedge are aggregated to generate the hyperedge embeddings.…”

Section: Methodsmentioning

confidence: 99%

Unveiling Tissue Structure and Tumor Microenvironment from Spatially Resolved Transcriptomics by Hypergraph Learning

Liao,

Zhang,

Wang

et al. 2024

Preprint

View full text Add to dashboard Cite

Spatially resolved transcriptomics (SRT) technologies acquire gene expressions and spatial information simultaneously, reshaping the perspectives of life sciences. Identifying spatial patterns is essential for exploring organ development and tumor microenvironment. Nevertheless, emerging SRT technologies have also introduced diverse spatial resolutions, posing challenges in characterizing spatial domains with finer resolutions. Here we propose a hypergraph-based method, termed HyperSTAR to precisely recognize spatial domains across varying spatial resolutions by utilizing higher-order relationships among spatially adjacent tissue programs. Specifically, a gene expression-guided hyperedge decomposition module is incorporated to refine the structure of the hypergraph to precisely delineate the boundaries of spatial domains. A hypergraph attention convolutional neural network is designed to adaptively learn the significance of each hyperedge. With the power of capturing intricate higher-order relationships within spatially neighboring multi-spots/cells, HyperSTAR demonstrates superior performance across different technologies with various resolutions compared to existing advanced graph neural network models in multiple tasks including uncovering tissue sub-structure, inferring spatiotemporal patterns, and denoising spatially resolved gene expressions. It successfully reveals spatial heterogeneity in breast cancer section and its findings are further validated through functional and survival analyses of independent clinical data. Notably, HyperSTAR performs well with diverse spatial omics data types and seamlessly extends to large-scale datasets.

show abstract

Section: Methodsmentioning

confidence: 99%

Unveiling Tissue Structure and Tumor Microenvironment from Spatially Resolved Transcriptomics by Hypergraph Learning

Liao,

Zhang,

Wang

et al. 2024

Preprint

View full text Add to dashboard Cite

show abstract

“…Examples where multiway relations cannot be reduced to an ensemble of pairwise relations are gene regulatory networks [19,20], where some reactions occur only when a set of multiple (not only two) genes interact, or applications in network neuroscience [21,22]. To overcome this problem, many architectures defined on general hypergraphs have been proposed [23][24][25][26][27], and some works incorporating attention mechanisms for hypergraphs neural networks have been published [28,29].…”

Section: Introductionmentioning

confidence: 99%

Cell Attention Networks

Giusti¹,

Battiloro²,

Testa³

et al. 2022

Preprint

View full text Add to dashboard Cite

Since their introduction, graph attention networks achieved outstanding results in graph representation learning tasks. However, these networks consider only pairwise relationships among nodes and then they are not able to fully exploit higher-order interactions present in many real world data-sets. In this paper, we introduce Cell Attention Networks (CANs), a neural architecture operating on data defined over the vertices of a graph, representing the graph as the 1skeleton of a cell complex introduced to capture higher order interactions. In particular, we exploit the lower and upper neighborhoods, as encoded in the cell complex, to design two independent masked self-attention mechanisms, thus generalizing the conventional graph attention strategy. The approach used in CANs is hierarchical and it incorporates the following steps: i) a lifting algorithm that learns edge features from node features; ii) a cell attention mechanism to find the optimal combination of edge features over both lower and upper neighbors; iii) a hierarchical edge pooling mechanism to extract a compact meaningful set of features. The experimental results show that CAN is a low complexity strategy that compares favorably with state of the art results on graph-based learning tasks.Preprint. Preliminary work.

show abstract

“…Such methods require complicated transformation processes and are prone to lose high-order information. The other is to transform and propagate vertex (node) feature information directly on hypergraph [9]- [11], [13], [15], [19], [24]- [26]. Vertex-hyperedge-vertex transformation is the most common feature transformation and propagation mode in this kind of networks.…”

mentioning

confidence: 99%

“…Different from HGNN, it uses multi-layer perception (MLP) based vertex convolution and hyperedge convolution to propagate features. By introducing attention mechanism [11], [25], [26], this transformation is further improved that can automatically learn the importance of vertices and hyperedges. The vertex-hyperedgevertex transformation is essentially weighted aggregation of neighborhood information for each node, whose neighborhood size is determined by its associated hyperedges, as shown in Fig.…”

mentioning

confidence: 99%

“…In addition, concerning the multi-modal representations in practice, how to effectively combine the multi-modal data is still a challenge. In the single hypergraph neural network models like [9]- [11], a hypergraph is constructed by concatenating hyperedges from features belonging to different modalities. In these methods, all modalities are treated equally, which is not able to reflect the importance of different modalities related to the task.…”

mentioning

confidence: 99%

See 1 more Smart Citation

Adaptive Multi-Hypergraph Convolutional Networks for 3D Object Classification

Nong

Peng

Zhang

et al. 2023

IEEE Trans. Multimedia

View full text Add to dashboard Cite

3D object classification is an important task in computer vision. In order to explore the high-order and multimodal correlations among 3D data, we propose an adaptive multi-hypergraph convolutional networks (AMHCN) framework to enhance 3D object classification performance. The proposed network improves the current hypergraph neural networks in two aspects. Firstly, existing networks rely on hyperedge constrained neighborhoods for feature aggregation, which may introduce noise or ignore positive information outside the hyperedges. To this end, we develop the partially absorbing random walks (PARW) to hypergraph for capturing optimal vertex neighborhoods from hypergraph globally. Then, based on the PARW on hypergraph, we design a new hypergraph convolution operator to learn deep embeddings from the optimized high-order correlation, which enables effective information propagation among the most relevant vertices. Secondly, concerning the multi-modal representations in practice, the current multi-modal hypergraph learning models either treat all modalities equally or introduce abundant parameters to learn weights of different modalities. To overcome these shortcomings, we propose a simple but effective dynamic weighting strategy for combining multi-modal representations, in which the importance of each modality can be adjusted adaptively by the loss function. We apply the proposed model to 3D object classification, and the experimental results on two 3D benchmark datasets demonstrate that our method outperforms the state-of-the-art methods, testifying to the effectiveness of both our convolution method and multi-modality fusion strategy.

show abstract

Hypergraph Attention Networks

Cited by 15 publications

References 20 publications

Unveiling Tissue Structure and Tumor Microenvironment from Spatially Resolved Transcriptomics by Hypergraph Learning

Unveiling Tissue Structure and Tumor Microenvironment from Spatially Resolved Transcriptomics by Hypergraph Learning

Cell Attention Networks

Adaptive Multi-Hypergraph Convolutional Networks for 3D Object Classification

Contact Info

Product

Resources

About