.png){kind=icon}
返回
Efficient and effective ensemble broad learning system based on structural diversity
.png){kind=icon}
PRE
AI
DOI：10.1016/j.asoc.2024.112412.png)
.png){kind=icon}
原文求助.png){kind=icon}
分享.png){kind=icon}
收藏
摘要
En 中文
Ensemble Broad Learning Systems (ENBLS) have received considerable attention from researchers and experienced rapid development due to its outstanding performance in handling complex data. A study has revealed that when ENBLS is trained with multiple independent storages, the issue of memory consumption is inevitable. To address this issue, this paper proposes an Ensemble Broad Learning Systems based on structural diversity (EBLSSD). This method effectively reduces the model memory consumption while maintaining superior performance. In EBLSSD, the scaling vectors are trained by constraining the scaling vectors of the current model and those of other existing models. These vectors are used to search for hidden paths of subnetworks within the network, extracting different network structures for integration. By preserving the scaling vectors during the storage process, memory pressure is reduced. EBLSSD introduces diversity from the perspective of network structure, which reduces cooperation between nodes and the risk of model overfitting, thereby enhancing model performance. The feasibility and effectiveness of the proposed algorithm is validated by experimental results on public datasets.
Keyword:
Broad learning system
Ensemble learning
Structural diversity
Scaling vector
期刊
IF:
6.6
论文数:
1.4W
被引数:
4.8W
机构
暂无机构信息
引用论文
An ensemble learning framework for convolutional neural network based on multiple classifiers.png)
SOFT COMPUTING
SOFT COMPUTINGIF2.5
PREAI
分享收藏LPI-BLS: Predicting lncRNA-protein interactions with a broad learning system-based stacked ensemble classifierLpi-bls: 使用基于广泛学习系统的堆叠集成分类器预测lncRNA-蛋白质相互作用NEUROCOMPUTING
NEUROCOMPUTINGIF6.5
PREAI
分享收藏分享收藏
查看更多内容