[1]王 梅,闫祖嘉,高雅田,等.基于自适应多头超图卷积网络的小样本回归模型[J].郑州大学学报(工学版),2026,47(5):85-92.[doi:10.13705/j.issn.1671-6833.2026.02.003]
　WANG Mei,YAN Zujia,GAO Yatian,et al.Few-shot Regression Model Based on Adaptive Multi-head Hypergraph Convolutional Network[J].Journal of Zhengzhou University (Engineering Science),2026,47(5):85-92.[doi:10.13705/j.issn.1671-6833.2026.02.003]
点击复制

基于自适应多头超图卷积网络的小样本回归模型()
分享到：

《郑州大学学报(工学版)》[ISSN:1671-6833/CN:41-1339/T]

卷:
47
期数:
2026年5期
页码:
85-92
栏目:
出版日期:
2026-09-09

文章信息/Info

Title:
Few-shot Regression Model Based on Adaptive Multi-head Hypergraph Convolutional Network
文章编号:
1671-6833(2026)05-0085-08
作者:
王 梅1,2, 闫祖嘉1, 高雅田1,2, 高俊涛1,2
1. 东北石油大学 计算机与信息技术学院, 黑龙江 大庆 163318;2. 东北石油大学 黑龙江省石油大数据与智能分析重点实验室,黑龙江 大庆 163318
Author(s):
WANG Mei1,2, YAN Zujia1, GAO Yatian1,2, GAO Juntao1,2
1. School of Computer and Information Technology, Northeast Petroleum University, Daqing 163318, China; 2. Heilongjiang Provincial Key Laboratory of Petroleum Big Data and Intelligent Analysis, Northeast Petroleum University, Daqing 163318, China
关键词:
小样本学习;  超图;  超图卷积神经网络;  多头注意力机制;  元学习
Keywords:
few-shot learning;  hypergraph;  hypergraph convolutional neural network;  multi head attention mechanism;  meta-learning
分类号:
TP301
DOI:
10.13705/j.issn.1671-6833.2026.02.003
文献标志码:
A
摘要:
针对图神经网络基于二元边结构难以捕捉多节点间的高阶交互，并且固定拓扑的图结构无法适应动态数据分布等问题，提出一种基于自适应多头超图卷积网络的小样本回归模型（AM‑HGCN）。首先，通过动态超图构建方法融合特征相似度与拓扑结构，结合k跳邻居(k‑hop)和k近邻(k\_NN)策略生成多尺度超边，自适应捕捉特征间的交互关系；其次，设计多头超图卷积网络，利用并行注意力头提取异构特征，并通过动态门控机制融合多粒度信息，增强模型的表达能力；最后，引入模型无关元学习框架，通过内外循环优化实现快速任务适应。在数据集Boston Housing、Energy Efficiency、IMDB、MinilmageNet上的实验表明：对于结构化数据集，AM‑HGCN在评价指标上显著优于主流基线模型，其中决定系数最高为0.775，验证了其对复杂关系建模的有效性。消融实验进一步证明，动态超图与多头注意力机制的协同作用是小样本回归性能提升的关键，实验结果验证了所提方法的有效性。
Abstract:
It was difficult to capture high‑order interactions among multiple nodes by graph neural networks based on binary‑edge structures, and static graph topologies were not adaptable to dynamic data distributions. To address these limitations, a few‑shot regression model based on adaptive multi‑head hypergraph convolutional networks (AM‑HGCN) was proposed. In this model, feature similarity and topological structure were integrated through a dynamic hypergraph construction method, multi‑scale hyperedges were generated using k\_hop neighbors and k-nearest neighbors (k\_NN) strategies to enable adaptive capture of feature interactions. Then, a multi‑head hypergraph convolutional network was designed to extract heterogeneous features via parallel attention heads and fuse multi‑granularity information through a dynamic gating mechanism to enhance expressive capability. Finally, a model‑agnostic meta‑learning framework was introduced to achieve rapid task adaptation through inner‑ and outer‑loop optimization. Experiments were conducted on the Boston Housing, Energy Efficiency, IMDB, and MinilmageNet datasets. For structured datasets, AM‑HGCN outperform mainstream baseline models significantly in evaluation metrics, The coefficient of determination was 0.775, which verifies its effectiveness in modeling complex relationships. Ablation studies further demonstrated that the collaborative effect of dynamic hypergraphs and multi‑head attention mechanisms was crucial for the enhancement of few‑shot regression performance, overall validating the effectiveness of the proposed method.

参考文献/References:

[1] Zhao Kailin, Jin Xiaolong, Wang Yuanzhuo. Survey on few‑shot learning[J]. Journal of Software, 2021, 32(2): 349‑369.［赵凯琳,靳小龙,王元卓.小样本学习研究综述[J].软件学报,2021,32(2):349‑369.］
[2] Wang Benben, Yan Yuefei, Lu Huizhen, et al. A predictive approach for cerebral small vessel disease based on imbalanced limited‑sample data[J]. Electro‑Mechanical Engineering, 2025, 41(3): 87‑92.［王奔犇,严粤飞,陆慧珍,等.基于小样本不平衡数据的脑小血管病预测方法[J].电子机械工程,2025,41(3):87‑92.］
[3] Zhou Wei, Wei Minggan, Xu Haixia, et al. A few‑shot land cover classification model for remote sensing images based on multimodality[J]. Journal of Electronics & Information Technology, 2025, 47(6): 1747‑1761.［周维,魏名安,许海霞,等.基于多模态的小样本遥感地物分类模型[J].电子与信息学报,2025,47(6):1747‑1761.］
[4] Gidaris S, Komodakis N. Generating classification weights with GNN denoising autoencoders for few‑shot learning[C]//Proceedings of the 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Piscataway: IEEE, 2019: 21‑30.
[5] Velickovic P, Cucurull G, Casanova A, et al. Graph attention networks[PP/OL]. V1. arXiv (2017‑10‑30)[2025‑10‑04]. https://doi.org/10.48550/arXiv.1710.10903.
[6] Zhao Xiao, Yun Xiaosa, Liu Ruiling, et al. Small sample plant image classification based on wavelet multi‑frequency feature fusion enhancement[J]. Computer Engineering and Applications, 2026, 62(11): 284‑294.［赵晓,负潇洒,刘睿玲,等.基于小波多频特征融合增强的小样本植物图像分类[J].计算机工程与应用,2026,62(11):284‑294.］
[7] Ge Xiaosan, Zheng Mengmeng. Study on classification of few‑shot remote sensing images based on SER‑GNN[J]. Journal of Henan Polytechnic University (Natural Science), 2025, 44(5): 144‑151.［葛小三,郑猛猛.基于SER‑GNN的小样本遥感影像分类研究[J].河南理工大学学报(自然科学版),2025,44(5):144‑151.］
[8] Shen Yu, Wang Ruoxuan, Li Jiangcheng, et al. A small‑sample time series prediction model based on graph structure[J/OL]. Journal of Beihang University, 2025:1‑14(2025‑07‑18)[2025‑07‑29].
https://doi.org/10.13700/j.bh.1001‑5965.2025.0198.［沈瑜,王若暄,李江城,等.基于图结构的小样本时间序列预测模型[J/OL].北京航空航天大学学报,2025:1‑14(2025‑07‑18)[2025‑07‑29].https://doi.org/10.13700/j.bh.1001‑5965.2025.0198.］
[9] Liang Zhuomin, Bai Liang, Yang Xian, et al. Multi‑channel disentangled graph neural networks with different types of self‑constraints[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2025, 47(9): 8001‑8012.
[10] Berge C. Hypergraphs: combinatorics of finite sets[M]. New York: Elsevier Science & Technology Books, 1984.
[11] Feng Yifan, You Haoxuan, Zhang Zizhao, et al. Hypergraph neural networks[J]. Proceedings of the AAAI Conference on Artificial Intelligence, 2019, 33(1): 3558‑3565.
[12] Bai Song, Zhang Feihu, Torr P H S. Hypergraph convolution and hypergraph attention[J]. Pattern Recognition, 2021, 110: 107637.
[13] Wu Yue, Wang Ying, Wang Xin, et al. Motif‑based hypergraph convolution network for semi‑supervised classification on heterogeneous graph[J]. Chinese Journal of Computers, 2021, 44(11): 2248‑2260.［吴越,王英,王鑫,等.基于超图卷积的异质网络半监督节点分类[J].计算机学报,2021,44(11):2248‑2260.］
[14] Jiang Jianwen, Wei Yuxuan, Feng Yifan, et al. Dynamic hypergraph neural networks[C]//Proceedings of the Twenty‑Eighth International Joint Conference on Artificial Intelligence. International Joint Conferences on Artificial Intelligence Organization, 2019: 2635‑2641.
[15] Feng Yifan, You Haoxuan, Zhang Zizhao, et al. Hypergraph neural networks[J]. Proceedings of the AAAI Conference on Artificial Intelligence, 2019, 33(1): 3558‑3565.
[16] Harrison D, Rubinfeld D L. Hedonic housing prices and the demand for clean air[J]. Journal of Environmental Economics and Management, 1978, 5(1): 81‑102.
[17] Chen Jiandong, Liu Jialu, Qi Jie, et al. City‑ and county‑level spatio‑temporal energy consumption and efficiency datasets for China from 1997 to 2017[J]. Scientific Data, 2022, 9: 101.
[18] Zhai Mengxin, Zhou Yanling, Yu Hang. Few‑shot text classification based on question‑oriented prompt‑tuning[J]. Application Research of Computers, 2025, 42(3): 708‑713.［翟梦鑫,周艳玲,余杭.基于问题导向式提示调优小样本文本分类[J].计算机应用研究,2025,42(3):708‑713.］
[19] Zhou Bin, Xiao Hao, Qin Yijia, et al. Local prototype small sample classification model integrated mixed attention mechanism[J]. Journal of Frontiers of Computer Science and Technology, 2026, 20(2): 584‑595.［周彬,鲜浩,秦艺嘉,等.融合混合注意力机制的局部原型小样本分类模型[J].计算机科学与探索,2026,20(2):584‑595.］
[20] Kipf T N, Welling M. Semi‑supervised classification with graph convolutional networks[PP/OL]. V4. arXiv(2017‑02‑22)[2025‑10‑04]. https://doi.org/10.48550/arXiv.1609.02907.
[21] Li Peiwen, Li Feijiang, Wang Jieting, et al. Clustering method for tabular data based on pretrained foundation models with synthetic data[J]. Journal of Computer Research and Development, 2025, 62(9): 2139‑2151.［李培文,李飞江,王婕婷,等.基于合成数据预训练基础模型的表格数据聚类方法[J].计算机研究与发展,2025,62(9):2139‑2151.］

更新日期/Last Update: 2026-09-07