.png){kind=icon}
.png){kind=icon}
Return
81-norm simultaneous approximate diagonalization and a proximal alternating maximization method
.png){kind=icon}
PRE
AI
DOI：10.1016/j.cam.2026.117421.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
This study investigates the 81-norm simultaneous approximate matrix diagonalization problem, aiming to mitigate the sensitivity of the 82-norm model to outliers. A proximal alternating maximization algorithm 81-PAM is developed using polar decomposition. We establish linear convergence based on the Euclidean Kurdyka-& Lstrok;ojasiewicz property and exponent. The algorithm and global convergence results are successfully extended to d-order tensors. Comparative experiments and color face reconstruction demonstrate the superior efficiency and robustness of 81-PAM against outliers and noises.
Keywords:
Simultaneous tensor/matrix diagonalization
Robust
Proximal alternating maximization
Linear convergence
Kurdyka-& Lstrok
ojasiewicz property
Journal
J
IF:
2.6
Papers:
336
Citations:
0