Optics & Photonics logo

Prof. Qing Qu uses data and machine learning to optimize the world

Qu’s research has applications in imaging sciences, scientific discovery, healthcare, and more.

Qing Qu headshot Enlarge

Prof. Qing Qu is a new faculty member in Electrical and Computer Engineering who specializes in Signal & Image Processing and Machine Learning. His research has applications in imaging sciences, scientific discovery, healthcare, and more.

In this Q&A, we chatted with Qu about his research, what he looks for in graduate students, why he chose Michigan, and his favorite hobbies, including cooking authentic Chinese food and his hope to adopt a corgi one day.

Tell us a bit about your current research. What sort of problems are you looking to address?

Broadly speaking, my research interest lies in the intersection of signal processing, data science, machine learning, and numerical optimization. I am particularly interested in developing guaranteed computational methods for learning low-complexity models from high-dimensional data, leveraging tools from machine learning, numerical optimization, and high dimensional geometry, with applications in imaging sciences, scientific discovery, and healthcare, etc.

In particular, my past work developed a new framework for studying nonconvex optimization problems in signal processing and machine learning, such as phase retrieval and dictionary learning. We showed that these challenging nonconvex problems often have benign global landscapes due to symmetry and low-dimensional structures, leading to the development of efficient global optimization methods. Very recently, I got very interested in understanding deep networks through the lens of landscape analysis and low-dimensional modeling, building upon the work we have done in the past.

Has the pandemic affected your ability to do this research? If so, how did you adapt?

Not necessarily. Because my research is more computation and theory-oriented, all I need are computers to run simulations to verify my conjectures, and a whiteboard to write the equations. Moreover, the advances of technology such as Zoom makes communication quite efficient. I do not feel much difference between working from the office and home. However, in the long run, I do think that in-person communication is very important for exchanging ideas; that cannot be replaced by something like Zoom. I did find it a bit hard to focus at the beginning, but I got used to it after a year.

What got you interested in this branch of engineering?

The interest is rooted in my family; my father is a geophysical engineer who claims himself “doing CT imaging for the earth.” During high school, I loved math and physics, so during my undergraduate career in Tsinghua I decided to join the academic talent program with a focus on math and physics. During my senior year, I felt more excited to use math and physics to change the world, so I changed my major to electrical engineering with a focus on signal processing. During my graduate study, I found that machine learning is rapidly transforming every discipline of engineering and science, and I was very convinced that this is the area that I can make the biggest impact.

Why did you choose Michigan? What excites you most about working here?

There are some obvious facts that made me excited to join U-M. The engineering school is consistently ranking top 5, and the ECE program is also among the top 5 in the nation. But more importantly, I think what made up my mind is more due to the people and collegial environment here. During the interview, I could feel that the faculty here really appreciated my work and were very supportive of me. I have a feeling that my research fits well into the current research in the area of signal processing and machine learning at U-M, and that I can find many collaborations within my area and across the engineering school.

Moreover, because of the top quality of the faculty and the university, it is easy to attract top students into my research group, building a high-quality research program into the future. All these factors made me really excited to join U-M and go BLUE!

What do you enjoy most about teaching?

I think one of the things that makes me feel really proud is when students tell me that they learned a lot from my class. I also enjoy integrating my research into teaching, because I believe the ultimate goal of the research is to gain knowledge and deliver the knowledge.

What qualities do you look for when selecting students to work with on research?

I have worked with quite a few master’s students during my Ph.D. at Columbia and my postdoc period at NYU. Some of them went to top Ph.D. graduate programs with my guidance. Based on my own limited experiences, I would say the qualities distinguishing top PhD students from others are their curiosity, strong motivation, and passion for research. I would assess them upon failures during research projects (which is quite common); good students will persist, learn and grow, eventually finding a new way out, while others might simply give up. I would also put academic integrity as a very important quality, because maintaining a high research quality and reputation is very important in the long run. I would put less emphasis on the overall GPA, where I found some students are good at research but don’t do that well in classes (unfortunately, I was one of them during my undergraduate career).

What are some of your favorite hobbies? Do you have any pets?

I spend most of my time on research, but I have some small hobbies. I spend one-hour per day doing exercises to keep myself fit and energetic. I do a variety of exercises, such as jogging, strength training, and swimming, etc. I really enjoyed jogging along the Hudson River when I was living in NYC. Another hobby that I enjoy is cooking during the weekend. I cook Chinese food that my friends appreciate a lot. I also like traveling around the world and visiting many different places. I do not have any pet, but plan to get a dog (maybe corgi?) if I have time!

Explore:
Big Data; Division News; Machine Learning; Profile; Qing Qu; Signal & Image Processing and Machine Learning