计算机与数字工程2025，Vol.53Issue(3)：786-794,9.DOI:10.3969/j.issn.1672-9722.2025.03.030
基于RVFL神经网络的集成深度学习
Ensemble Deep Learning Based on RVFL Neural Network
摘要
Abstract
This paper proposes a deep learning framework based on stochastic neural network.Inspired by the principle of Random Vector Functional Link(RVFL)network,a deep RVFL network with stacked layers(dRVFL)is proposed.The network can be used as a weighted network while extracting the features of multiple hidden layers so as to provide weights for the original features obtained by direct links and the features of hidden layers.Secondly,an integrated deep RVFL network(edRVFL)combining inte-grated learning and deep learning is proposed.Compared to traditional integration methods,edRVFL only trains a single dRVFL net-work once.The closed-form solution in the network greatly reduces the non-iterative training time.At the same time,the frame-works for both networks can be used with any RVFL variant.Finally,the proposed network is compared with nine different deep feedforward neural networks on 46 tabular UCI classification datasets and 12 sparse datasets,and the results show that the proposed network has better performance.关键词
RVFL/深度学习/集成学习/深度前馈神经网络Key words
RVFL/deep learning/ensemble learning/deep feedforward neural network分类
信息技术与安全科学引用本文复制引用
郭金成,魏霖静..基于RVFL神经网络的集成深度学习[J].计算机与数字工程,2025,53(3):786-794,9.基金项目
2020年甘肃农业大学研究生教育研究项目(编号:2020-19) （编号:2020-19）
2021年度兰州市人才创新创业项目(编号:2021-RC-47) （编号:2021-RC-47）
2021年教育部产学研合作协同育人项目(编号:202102326036)资助. （编号:202102326036）
