Multimodal Recommendation System Based on Cross Self-Attention Fusion
Recent advances in graph neural networks (GNNs) have enhanced multimodal recommendation systems’ ability to process complex user–item interactions. However, current approaches face two key limitations: they rely on static similarity metrics for product relationship graphs and they struggle to effect...
Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2025-01-01
|
| Series: | Systems |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2079-8954/13/1/57 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1832587427454124032 |
|---|---|
| author | Peishan Li Weixiao Zhan Lutao Gao Shuran Wang Linnan Yang |
| author_facet | Peishan Li Weixiao Zhan Lutao Gao Shuran Wang Linnan Yang |
| author_sort | Peishan Li |
| collection | DOAJ |
| description | Recent advances in graph neural networks (GNNs) have enhanced multimodal recommendation systems’ ability to process complex user–item interactions. However, current approaches face two key limitations: they rely on static similarity metrics for product relationship graphs and they struggle to effectively fuse information across modalities. We propose MR-CSAF, a novel multimodal recommendation algorithm using cross-self-attention fusion. Building on FREEDOM, our approach introduces an adaptive modality selector that dynamically weights each modality’s contribution to product similarity, enabling more accurate product relationship graphs and optimized modality representations. We employ a cross-self-attention mechanism to facilitate both inter- and intra-modal information transfer, while using graph convolution to incorporate updated features into item and product modal representations. Experimental results on three public datasets demonstrate MR-CSAF outperforms eight baseline methods, validating its effectiveness in providing personalized recommendations, advancing the field of personalized recommendation in complex multimodal environments. |
| format | Article |
| id | doaj-art-61df0415a6d54eef9b78ba00d6f08cf3 |
| institution | Kabale University |
| issn | 2079-8954 |
| language | English |
| publishDate | 2025-01-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Systems |
| spelling | doaj-art-61df0415a6d54eef9b78ba00d6f08cf32025-01-24T13:50:39ZengMDPI AGSystems2079-89542025-01-011315710.3390/systems13010057Multimodal Recommendation System Based on Cross Self-Attention FusionPeishan Li0Weixiao Zhan1Lutao Gao2Shuran Wang3Linnan Yang4College of Big Data, Yunnan Agricultural University, Kunming 650201, ChinaCollege of Computer Science and Engineering, University of California, San Diego, CA 92093, USACollege of Big Data, Yunnan Agricultural University, Kunming 650201, ChinaCollege of Big Data, Yunnan Agricultural University, Kunming 650201, ChinaCollege of Big Data, Yunnan Agricultural University, Kunming 650201, ChinaRecent advances in graph neural networks (GNNs) have enhanced multimodal recommendation systems’ ability to process complex user–item interactions. However, current approaches face two key limitations: they rely on static similarity metrics for product relationship graphs and they struggle to effectively fuse information across modalities. We propose MR-CSAF, a novel multimodal recommendation algorithm using cross-self-attention fusion. Building on FREEDOM, our approach introduces an adaptive modality selector that dynamically weights each modality’s contribution to product similarity, enabling more accurate product relationship graphs and optimized modality representations. We employ a cross-self-attention mechanism to facilitate both inter- and intra-modal information transfer, while using graph convolution to incorporate updated features into item and product modal representations. Experimental results on three public datasets demonstrate MR-CSAF outperforms eight baseline methods, validating its effectiveness in providing personalized recommendations, advancing the field of personalized recommendation in complex multimodal environments.https://www.mdpi.com/2079-8954/13/1/57graph neural networksmultimodal recommendation systemsattention mechanismpersonalized recommendation |
| spellingShingle | Peishan Li Weixiao Zhan Lutao Gao Shuran Wang Linnan Yang Multimodal Recommendation System Based on Cross Self-Attention Fusion Systems graph neural networks multimodal recommendation systems attention mechanism personalized recommendation |
| title | Multimodal Recommendation System Based on Cross Self-Attention Fusion |
| title_full | Multimodal Recommendation System Based on Cross Self-Attention Fusion |
| title_fullStr | Multimodal Recommendation System Based on Cross Self-Attention Fusion |
| title_full_unstemmed | Multimodal Recommendation System Based on Cross Self-Attention Fusion |
| title_short | Multimodal Recommendation System Based on Cross Self-Attention Fusion |
| title_sort | multimodal recommendation system based on cross self attention fusion |
| topic | graph neural networks multimodal recommendation systems attention mechanism personalized recommendation |
| url | https://www.mdpi.com/2079-8954/13/1/57 |
| work_keys_str_mv | AT peishanli multimodalrecommendationsystembasedoncrossselfattentionfusion AT weixiaozhan multimodalrecommendationsystembasedoncrossselfattentionfusion AT lutaogao multimodalrecommendationsystembasedoncrossselfattentionfusion AT shuranwang multimodalrecommendationsystembasedoncrossselfattentionfusion AT linnanyang multimodalrecommendationsystembasedoncrossselfattentionfusion |