ISBN: 978-981-18-3959-7 DOI: 10.18178/wcse.2022.06.016
Missing Data Imputation Approach Based on Adaptive Compressed Sensing
Abstract— The paper presents a missing data imputation method based on compressed Sensing (CS). First of all, the problem of data imputation is translated into the recovery of sparse vector under the framework of compressed sensing. Secondly, we propose an improved greedy reconstruction algorithm called Double Try Sparsity Adaptive Matching Pursuit (DTSAMP). The algorithm obtains the estimation of sparsity by trying twice to approximate the value of sparsity, and then approximates the estimated value in each iteration. As a result, the missing data sets can be reconstructed without prior information of the sparsity. Furthermore, the step size and support set are well controlled by setting thresholds during the iteration. The simulation results show that the proposed algorithm is superior to other methods in terms of reconstruction speed and accuracy, as well as better robustness.
Index Terms—data imputation, compressed sensing, sparsity adaptive, step length control, maching pursuit
Bing Ren, Yan Guo, Duan Xue, Zhijian Li
College of Communication Engineering, Army Engineering University of PLA, CHINA
Departments of Information and Communication, National University of Defense Technology, CHINA
Cite: Bing Ren, Yan Guo, Duan Xue, Zhijian Li, Zhigang Jia, "Missing Data Imputation Approach Based on Adaptive Compressed Sensing, " Proceedings of 2022 the 12th International Workshop on Computer Science and Engineering (WCSE 2022), pp. 103-111, June 24-27, 2022.