BLSHF: Broad Learning System with Hybrid Features

Cao，Weipeng1,2; Li，Dachuan3; Zhang，Xingjian1; Qiu，Meikang4; Liu，Ye2

10.1007/978-3-031-10986-7_53

2022

15th International Conference on Knowledge Science, Engineering, and Management (KSEM)

0302-9743

1611-3349

978-3-031-10985-0

LECT NOTES ARTIF INT

13369 LNAI

655-666

AUG 06-08, 2022

null,Singapore,SINGAPORE

GEWERBESTRASSE 11, CHAM, CH-6330, SWITZERLAND

SPRINGER INTERNATIONAL PUBLISHING AG

Broad Learning System (BLS), a type of neural network with a non-iterative training mechanism and adaptive network structure, has attracted much attention in recent years. In BLS, since the mapped features are obtained by mapping the training data based on a set of random weights, their quality is unstable, which in turn leads to the instability of the generalization ability of the model. To improve the diversity and stability of mapped features in BLS, we propose the BLS with Hybrid Features (BLSHF) algorithm in this study. Unlike original BLS, which uses a single uniform distribution to assign random values for the input weights of mapped feature nodes, BLSHF uses different distributions to initialize the mapped feature nodes in each group, thereby increasing the diversity of mapped features. This method enables BLSHF to extract high-level features from the original data better than the original BLS and further improves the feature extraction effect of the subsequent enhancement layer. Diverse features are beneficial to algorithms that use non-iterative training mechanisms, so BLSHF can achieve better generalization ability than BLS. We apply BLSHF to solve the problem of air quality evaluation, and the relevant experimental results empirically prove the effectiveness of this method. The learning mechanism of BLSHF can be easily applied to BLS and its variants to improve their generalization ability, which makes it have good application value.

Air quality prediction Broad Learning System Neural networks

Others

English

EI ; CPCI-S

National Natural Science Foundation of China[62106150] ; CAAC Key Laboratory of Civil Aviation Wide Surveillance and Safety Operation Management and Control Technology[202102] ; CCF-NSFOCUS[2021001]

Computer Science

Computer Science, Artificial Intelligence ; Computer Science, Information Systems ; Computer Science, Theory & Methods

WOS:000877369500053

20223112462282

Air quality ; Iterative methods

Air Pollution Control:451.2 ; Numerical Methods:921.6

2-s2.0-85135011131

Scopus

1.CAAC Key Laboratory of Civil Aviation Wide Surveillance and Safety Operation Management and Control Technology,Civil Aviation University of China,Tianjin,China
2.College of Computer Science and Software Engineering,Shenzhen University,Shenzhen,China
3.Department of Computer Science and Engineering,Southern University of Science and Technology,Shenzhen,China
4.Department of Computer Science,Texas A &M University-Commerce,Commerce,United States

Cao，Weipeng,Li，Dachuan,Zhang，Xingjian,et al. BLSHF: Broad Learning System with Hybrid Features[C]. GEWERBESTRASSE 11, CHAM, CH-6330, SWITZERLAND:SPRINGER INTERNATIONAL PUBLISHING AG,2022:655-666.