Semantic-aware heterogeneous information network embedding with incompatible meta-paths

Zheng, Susu; Guan, Donghai; Yuan, Weiwei

doi:10.1007/s11280-021-00903-5

“…To evaluate the performance of the proposed method, we consider both the stateof-the-art methods of recommendation and heterogeneous graph convolutional neural networks. They include ACKRec [15], a method combines meta-paths and graph neural networks to extract features of users and course concepts for course recommendation; MRCRec [16], a method of multi-entity relational self-symmetric meta-path for course recommendation; HAN [19], a heterogeneous graph neural network which uses node-level attention and semantic-level attention to obtain the weights of nodes and meta-paths respectively; ie-HGCN [20], a heterogeneous graph convolutional neural network that uses object-level aggregation and type-level aggregation to discover the best meta-path for nodes embeddings; SAHE [21], a method aggregate node similarity relationships to obtain the node similarity and use KL Divergence to learn node embedding to solve the meta-path incompatibility problem; HetSANN [22], a heterogeneous graph neural network for mining semantic information in heterogeneous networks.…”

Section: Evaluation Metrics and Implementation Detailsmentioning

confidence: 99%

“…Message passing importance of all possible meta-paths. SAHE [21] minimized the distance between the aggregated similarity matrix and the meta-path based similarity matrices to solve the incompatibility problem. Recently, some methods perform heterogeneous graph embedding learning without the use of pre-defined meta-paths.…”

Section: Message Passingmentioning

confidence: 99%

Adaptive Community Aware Graph Network for Item Recommendation

Hao

¹

,

Qian

²

,

Wang

³

et al. 2023

Preprint

0

View full text Add to dashboard Cite

Due to the heterogeneity of a large amount of real-world data, meta-paths are widely used in recommendation. Such recommendation methods can represent composite relationships between entities, but cannot explore reliable relations between nodes and influence among meta-paths. For solving this problem, adaptive community aware graph network (ACAGN) is proposed in this paper for item recommendation. By dynamically adapting the shape of communities and simultaneously calculating the similarity between them, the similarity of communities is fully learnt among nodes in meta-path based subgraphs. The semantic features of users and items are used to jointly learn weights between nodes, further to get reliable node relations, resulting in robust embedding. Finally, the embeddings of users and items are fed to extend matrix factorization for getting the top recommendations. A series of comprehensive experiments are conducted on two different public datasets. The results show that ACAGN is a better way to do item recommendation compared with the state-of-the-art methods.

show abstract

“…Yun et al [47] developed GTN to automatically identify useful graph connections. A technique dubbed MAHINDER is proposed in [55,57] to employ and encode meta-paths over different views with attention on the importance of attributes and data views. In an unsupervised setting, a heterogeneous graph neural network is proposed in [51], which samples a fixed size of neighbours and fuses their features using LSTMs.…”

Section: Heterogeneous Graph Neural Networkmentioning

confidence: 99%

GoMIC: Multi-view Image Clustering via Self-supervised Contrastive Heterogeneous Graph Co-learning

Fang

¹

,

Li

²

,

Akhtar

³

et al. 2022

Preprint

1

0

View full text Add to dashboard Cite

Graph learning is being increasingly applied to image clustering to reveal intra-class and inter-class relationships in data. However, existing graph learning-based image clustering focuses on grouping images under a single view, which under-utilises the information provided by the data. To address that, we propose a self-supervised multi-view image clustering technique under contrastive heterogeneous graph learning. Our method computes a heterogeneous affinity graph for multi-view image data. It conducts Local Feature Propagation (LFP) for reasoning over the local neighbourhood of each node and executes an Influence-aware Feature Propagation (IFP) from each node to its influential node for learning the clustering intention. The proposed framework pioneeringly employs two contrastive objectives. The first targets to contrast and fuse multiple views for the overall LFP embedding, and the second maximises the mutual information between LFP and IFP representations. We conduct extensive experiments on the benchmark datasets for the problem, i.e. COIL-20, Caltech7 and CASIA-WebFace. Our evaluation shows that our method outperforms the state-of-the-art methods, including the popular techniques MVGL, MCGC and HeCo.

show abstract

“…[39] developed GTN to automatically identify useful graph connections. A technique dubbed MAHINDER is proposed in [40,41] to employ and encode meta-paths over different views with attention on the importance of attributes and data views. In an unsupervised setting, a heterogeneous graph neural network is proposed in [42], which samples a fixed size of neighbours and fuses their features using LSTMs [43].…”

Section: Heterogeneous Graph Neural Networkmentioning

confidence: 99%

GoMIC: Multi-view image clustering via self-supervised contrastive heterogeneous graph co-learning

Fang

¹

,

Li

²

,

Akhtar

³

et al. 2022

World Wide Web

23

0

View full text Add to dashboard Cite

Graph learning is being increasingly applied to image clustering to reveal intra-class and inter-class relationships in data. However, existing graph learning-based image clustering focuses on grouping images under a single view, which under-utilises the information provided by the data. To address that, we propose a self-supervised multi-view image clustering technique under contrastive heterogeneous graph learning. Our method computes a heterogeneous affinity graph for multi-view image data. It conducts Local Feature Propagation (LFP) for reasoning over the local neighbourhood of each node and executes an Influence-aware Feature Propagation (IFP) from each node to its influential node for learning the clustering intention. The proposed framework pioneeringly employs two contrastive objectives. The first targets to contrast and fuse multiple views for the overall LFP embedding, and the second maximises the mutual information between LFP and IFP representations. We conduct extensive experiments on the benchmark datasets for the problem, i.e. COIL-20, Caltech7 and CASIA-WebFace. Our evaluation shows that our method outperforms the state-of-the-art methods, including the popular techniques MVGL, MCGC and HeCo.

show abstract

Semantic-aware heterogeneous information network embedding with incompatible meta-paths

Cited by 18 publications

References 35 publications

Adaptive Community Aware Graph Network for Item Recommendation

Adaptive Community Aware Graph Network for Item Recommendation

GoMIC: Multi-view Image Clustering via Self-supervised Contrastive Heterogeneous Graph Co-learning

GoMIC: Multi-view image clustering via self-supervised contrastive heterogeneous graph co-learning

Contact Info

Product

Resources

About