Reweighted nuclear norm regularization: A SPARSEVA approach^*

详细信息    查看全文

• 作者：Huong Ha^* ; ^{huong.ha@uon.edu.au" class="auth_mail" title="E-mail the corresponding author} ; James S. Welsh^* ; ^{james.welsh@newcastle.edu.au" class="auth_mail" title="E-mail the corresponding author} ; Niclas Blomberg^** ; ^{nibl@kth.se" class="auth_mail" title="E-mail the corresponding author} ; Cristian R. Rojas^** ; ^{cristian.rojas@ee.kth.se" class="auth_mail" title="E-mail the corresponding author} ; Bo Wahlberg^** ; ^{bo.wahlberg@ee.kth.se" class="auth_mail" title="E-mail the corresponding author}
• 关键词：system identification ; discrete time system ; regularization ; re-weighted nuclear norm ; sparse estimate
• 刊名：IFAC-PapersOnLine
• 出版年：2015
• 出版时间：2015
• 年：2015
• 卷：48
• 期：28
• 页码：1172-1177
• 全文大小：471 K

文摘

The aim of this paper is to develop a method to estimate high order FIR and ARX models using least squares with re-weighted nuclear norm regularization. Typically, the choice of the tuning parameter in the reweighting scheme is computationally expensive, hence we propose the use of the SPARSEVA (SPARSe Estimation based on a VAlidation criterion) framework to overcome this problem. Furthermore, we suggest the use of the prediction error criterion (PEC) to select the tuning parameter in the SPARSEVA algorithm. Numerical examples demonstrate the veracity of this method which has close ties with the traditional technique of cross validation, but using much less computations.