Lecturer Profile |
Ningning Han received the B.S. degree from University of Jinan, and the M.S. and Ph. D. degrees from Tianjin University in 2014 and 2017 respectively. He was a visiting scholar in San Francisco State University from 2016 to 2017. Since 2017, He is a postdoctoral fellow at Hong Kong Baptist University. His current research interests include compressed sensing, image processing and signal sparse representation. |
Lecture Abstract |
Solving linear inverse problems is a fundamental problem in engineering and scientific fields with many real-word applications. A class of iterative thresholding algorithms based on null space tuning, hard thresholding, feedbacks (suboptimal feedbacks) is attractive due to its simplicity and thus is exceedingly effective for large system recovery problems. The key of this theory is the concept that the number of indices selected at each iteration should be considered in order to speed up the convergence. Our results show that uniform recovery of all s-sparse signals from given possibly linear measurements can be achieved under the certain preconditioned restricted isometry condition and restricted isometry condition and it is possible to accelerate the convergence speed and improve the convergence condition by selecting an appropriate size of the support per iteration. Particularly, the theory provides a new perspective for other iterative shrinkage-thresholding algorithms related to recent advances in sparse recovery and deep learning. The theoretical findings can be demonstrated by numerical experiments. |
