Work

Bilinear Compressed Sensing

Year: 2019
Type: dissertation
Abstract: This dissertation studies the performance of convex and non-convex algorithms for randomized bilinear inverse problems.In particular, we examine algorithmic approaches based on convex relaxation for certain models of the blind deconvolution problem.These models are especially relevant for wireless communications.Moreover, this thesis analyses whether the required number of measurements can be reduced when additional information about the signals is available such as sparsity. ZusammenfassungDiese Dissertation untersucht die Leistungsfähigkeit von konvexen und nicht-konvexen Algorithmen für randomisierte bilineare inverse Probleme.Insbesondere werden algorithmische Ansätze für bestimmte Modelle des Problems der blinden Entfaltung betrachtet, welche auf konvexer Relaxation basieren.Diese Modelle sind insbesondere für die Nachrichtentechnik von Bedeutung.Des Weiteren wird in dieser Arbeit untersucht, ob die Anzahl der benötigten Messungen reduziert werden kann, falls zusätzliche Informationen über die Signale vorhanden sind wie etwa deren Dünnbesetztheit.To begin with, I would like to express my gratitude to my advisor Professor Felix Krahmer, who gave me the opportunity to do research in this exciting field.His approach to research was ideal for me.While always insisting on tackling relevant and important problems, he nevertheless gave me enough freedom so that I could follow my interests and find my own path in mathematics.I learned a lot from his advice, which often went beyond purely scientific matters.Moreover, I have been very lucky that I got the opportunity to collaborate with Peter Jung from the very beginning of my doctoral studies.Whenever I had questions concerning research and beyond, he was willing to help.Moreover, I profited a lot from his ability to translate problems arising in communications into mathematical questions.Furthermore, I want to thank all my scientific collaborators as well as all the numerous people I had the opportunity to meet and discuss science with.I've learned so much from you!I would like to thank Professor Stefan Weltge for being chair of my defense.In addition, I am grateful to Professor Kiryung Lee and Professor Yaniv Plan for serving on my thesis committee.Furthermore, I would like to thank all my colleagues in the math department at TUM for creating such a pleasant environment.In particular, I want to thank all members of the research group M15 for all the nice discussions we had during the last years.Moreover, I was very fortunate to have great friends in Munich.I will never forget the nice weekends, where we went biking or where we enjoyed Thai (and German) food;-) Finally, I am grateful to my family for their love and their constant support through all the good and bad times of my life.Thank you!Then with probability at least 1 -O L -1 every minimizer X of the SDP (2.2) satisfies(2.3)The coherence parameters µ 2 max , µ 2 min , and µ 2 h 0 are defined by µ
Language: English
Open Access status: green
OpenAlex ID: W7017842059
Cites: 160
Cited by:
Related to: 9