.png){kind=icon}
Return
MoBLS: A broad learning system with multi-objective optimization
.png){kind=icon}
PRE
AI
DOI：10.1016/j.asoc.2025.114454.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Broad learning system (BLS) attracts extensive attention as an effective alternative to deep neural networks. However, the performance of BLS strongly depends on the proper weight selection. More importantly, the number of neurons has a significant impact on its performance. To address these issues, this paper proposes a broad learning system based on multi-objective optimization (MoBLS). In MoBLS, the optimization objectives include the approximation degree of the model and the sparsity of the weights. Thanks to the strongly convex sparse operator, the weights of the model are effectively sparsified. In addition, the application of a population competition algorithm that combines the global and winner information leads to powerful approximation capabilities. When encountering different numbers of neurons, the experimental results show that the generalization performance of the MoBLS model statistically outperforms the classic BLS model and its state-of-the-art variants.
Keywords:
Broad learning system
Multi-objective optimization
Strongly convex sparse
Population competition
Robust
Journal
IF:
6.6
Papers:
1.4W
Citations:
4.8W
Organization
View more organizations
Cited Papers
.png){kind=icon}
ShareSave
ShareSaveShareSaveShareSaveHybrid and enhanced PSO: Novel first order reliability method-based hybrid intelligent approaches
PREAI
ShareSaveShareSaveShareSave View more