Research Article
Adaptive Feature Selection Based on Low-Rank Representation
@INPROCEEDINGS{10.1007/978-3-031-04245-4_30, author={Ying Wang and Lijun Fu and Hongwei Zhao and Qiang Fu and Guangyao Zhai and Yutong Niu}, title={Adaptive Feature Selection Based on Low-Rank Representation}, proceedings={6GN for Future Wireless Networks. 4th EAI International Conference, 6GN 2021, Huizhou, China, October 30--31, 2021, Proceedings}, proceedings_a={6GN}, year={2022}, month={5}, keywords={Adaptive Feature selection Low-rank}, doi={10.1007/978-3-031-04245-4_30} }- Ying Wang
Lijun Fu
Hongwei Zhao
Qiang Fu
Guangyao Zhai
Yutong Niu
Year: 2022
Adaptive Feature Selection Based on Low-Rank Representation
6GN
Springer
DOI: 10.1007/978-3-031-04245-4_30
Abstract
In the existing feature selection methods, the ways of construct the similarity matrix is: the first way to construct is give fixed value to two data, and the second is to calculate the distance between the two data and use it as the similarity. However, the above-mentioned method of constructing a similarity matrix is usually unreliable because the original data is often affected by noise. In the article, an adaptive feature selection method based on low-rank representation was proposed. In the method, we would dynamically construct a similarity matrix with local adaptive capabilities based on the feature projection matrix learned by the method. This construction way can reduce the influence of noise on the similarity matrix. To verify the validity of the method, we test our method on different public data sets.
