Interview with Professor Katie Bouman

Interviewer: Agnim Agarwal

Trivia

  • Favorite book: I really like Gone with the Wind. I know that there are a lot of issues with it; it is important to understand that it is a heavily biased and, therefore, misleading portrayal of the South during the Civil War. However, I still love the story and the character development, so, growing up, I read this book many times.
  • Favorite place: My favorite place is being with my family, wherever they are. When I was younger, my grandparents lived by the shore in Southern New Jersey. We would always go to the boardwalk and the beach, and I have such fond memories of being there. Although I like to travel, I would be perfectly happy going to the middle of nowhere, as long as my family’s around and we can hang out.
  • Favorite food: I like pasta a lot. Recently, we got an attachment that allows us to make our own pasta. This past weekend, we made hand-pulled noodles with some friends. So, in addition to enjoying eating pasta, I’m enjoying learning how to make different kinds of pasta.
  • Favorite computational imaging technique: Going back to my undergraduate education, one topic that I remember being amazed by and finding so beautiful was sampling theory. I remember learning about how all the pieces fit together and thinking, “wow, this is so beautiful.” Sampling theory is not necessarily a computational imaging technique, but it is a theory that is behind a lot of the stuff that I do.

Can you provide an overview of your group’s research areas? What are some of the possible real-world applications? 

My group works on a lot of different problems in computational imaging and machine learning for inverse problems. Computational imaging is all about how to design new hardware and algorithms that work together to see or measure things we can’t see with traditional sensors.

However, as we continue to push farther into what we can design with our own ingenuity, I think it’s important to think about how we can use machine learning to help us design new kinds of imaging pipelines for the future. Designing AI to guide a scientist or engineer in what they should measure or using AI to design new types of cameras altogether is an area that I’m really excited about.

A lot of times in computational imaging pipelines, sensors produce data that is very sparse or noisy, very messy to interpret. So, another area that my group is very interested in is how do we incorporate ideas from domain knowledge into a problem to add structure, so we can better pull out information that is impossible to obtain otherwise. For instance, one of my students – Berthy Feng – is currently working in a collaboration in mechanical engineering with Prof. Chiara Daraio and her student Alex Ogren. Berthy is working on developing a technique to recover a picture of the inside of a vibrating object that is being observed with a single monocular camera. So here is an example of a project where we use domain knowledge with ideas from machine learning and optimization to recover something that’s invisible to us. Techniques like this may be helpful in the future for a variety of applications including non-destructive testing.

Many problems that my group is working on are applicable to other domains. For instance, I have a student who was working with colleagues in Caltech’s seismology department on problems relevant to localization of earthquakes and ground tomography. Another one of my students is working on developing ideas for more efficient MRI (magnetic resonance imaging); this would be important for many medical challenges, including better medical imaging of children that have a hard time laying perfectly still for long periods of time.  Additionally, members of my group are working on astronomical imaging applications. In particular, techniques for better imaging of black holes and algorithms that help us design new telescopes that will allow us to image black holes better in the future. For instance, one of my postdocs is working on a method to recover information about the dynamics of a black hole and how it’s evolving; this information may help us to learn more about black hole evolution and possibly even gravity.

My group works on a number of different domains, but the core idea that cuts across all problems is designing new computational pipelines that merge unique kinds of sensing with new algorithms to extract more information than we could previously.

What motivated you to enter the field of computational imaging and, on a broader level, embark on a career in scientific research?

I actually started as a SURF student doing research. I worked in an image and video processing laboratory. At the time, I was helping a graduate student with a project where the goal was to automatically identify what camera took a particular image. Basically, the technique leveraged small signals in the noise of images. In particular, every sensor on a digital camera has imperfections and those imperfections cause different noise distributions that appear in the picture you take. Although you can’t see the differences in the noise, if you look across many images, you might be able to pick up on these tiny little signatures. They act like fingerprints of the camera.

As I was helping with the project, I got really excited about this idea that images contain all this hidden information. And, again, if we know where to look for it, we can try to tease that information out of the image. This idea excited me. I also liked working with images because it was very visual, and I liked being able to visualize the results that we were getting. My interest in this initial project eventually led me to do a PhD in a computer vision group, which is all about understanding and analyzing images.

I didn’t really get into scientific applications until a few years later, when I was introduced to the Event Horizon Telescope project. I was very excited by the collaboration’s goal to take the first image of a black hole, and the problems they were tackling shared a lot of similarities to problems that I had worked on in the past: you want to use these tiny signatures to tell you about underlying properties of an unknown system. So, although black hole imaging was a very different problem than I had been working on before, it shared a lot of structure with problems I had worked on in the past. Since then, I have loved using computation, machine learning, and new ideas to help push scientific imaging.

Since coming to Caltech, I’ve branched out to different domains and applications. That’s one thing I love about Caltech actually — that it’s so easy to talk to people across different departments. It allows us to more easily develop new ideas with people from different areas.

What are some of the challenges you’ve recently faced, and what do you foresee as the future directions of your research?

One of the big things I want to do is to develop new computational tools that tell us how to be smart in what new cameras we design. A lot of times, ideas, like the Earth-sized telescope that took the first image of a black hole, are just dreamed up by people who think hard and are very creative. They think, “oh, if I do this, then maybe I can get this information out.” It’s been amazing, the progress that has been made by our own human ingenuity. However, I think that there are likely other kinds of imaging techniques or modalities that are still unknown to us. Restricting ourselves to our own creativity might be hurting us. I’d like to think about how we can use collections of data to help us learn new imaging pipelines that we haven’t yet dreamed up. That’s an exciting futuristic goal of mine, and we’re taking baby steps towards that goal now in my group.

Also, when you use deep machine learning architectures, oftentimes it’s very hard to understand why the machine makes certain decisions. So, another thing I’m interested in is how to interpret the machine and use this to train scientists in identifying what are good features in scientific data.

These future goals of mine may sound a little all over the place, but they all feed into the same theme of using machine learning and our domain knowledge together to try to build better computational cameras and recover more hidden information that’s out there.

How has your research and day-to-day life shifted as a result of the ongoing pandemic?

Too many zoom calls, but I’m amazed at what we’ve been able to do, even though we are all working at home. I’m so amazed by all the progress students are making on different kinds of research and homework, even though they don’t have peers physically around them. I know when I was a graduate student, so many of my ideas popped into my head when getting coffee with my other lab mates, or when tapping on a friend’s shoulder and saying “hey, I have this problem, can you talk it out with me for five minutes?” I’m so impressed by the students and how they’ve coped, despite not having these kinds of impromptu interactions. Overall, we’ve done much better than I thought we would. That being said, I do hope we’re back soon, and I won’t miss having countless zoom meetings—I definitely like the in-person meetings more.

What aspects of being a researcher do you most enjoy?

I’m really curious about a lot of different problems, and I especially love working on problems at the intersection of different domains, because it allows me to learn about a new area I didn’t know about before. Back to when I first started working on the black hole imaging project: when I started, I didn’t know anything about black holes, but my collaborators basically taught me everything I know about astrophysics and that’s been awesome. Similarly, here at Caltech, people in the seismology department – in particular, Zach Ross, who we’ve been collaborating with – have been great at teaching me new things. In general, I think learning about new areas is one big motivation for doing new types of research.

It’s exciting to pose a question that no one has ever answered or that has never been approached in a certain way. Seeing if you can come up with a new method, approach, or idea that pushes the bar forward – that to me is really exciting. It’s so different than doing homework where you know there is always an answer. Here, it’s always unknown. And most of the time, to be honest, nothing ever works. Most of the time, nothing’s working and you only get small doses of success. But I think that it’s pretty exciting to be able to work with people to create something new.

I also love working with all of the great students at Caltech; they’re awesome and that’s what really makes it worth it (even when things are not working as expected).

What interests do you have outside of science?

I like Southern California. It is great how we have sun every day, so I get to explore more of the outdoors. I love going on picnics. Another thing that I’ve been enjoying recently is getting together with a couple of friends and cooking new, different kinds of meals. Recently for Chinese New Year, we made dumplings for the first time and it’s been a lot of fun to try making new kinds of foods. Another hobby I’ve taken up recently is crocheting; it’s relaxing, and I made my first sweater, which was fun.

Who influences you most both personally and academically? Who are your role models?

I have lots of role models. There have been many people throughout my life who have encouraged me, and whom I consider to be my role models and mentors, ranging from my own peers to more senior people. I can’t pinpoint any one person, but my family has been particularly important. They have always encouraged me to keep pushing forward. I also had mentors back in my high school, I did a SURF project with some wonderful mentors, and then I had mentors throughout undergrad and an amazing PhD advisor. Now, I have a lot of great mentors at Caltech, so I’m lucky to have role models from many different places.

What advice would you give to Caltech students and other young students interested in research?

Find a topic that really excites you and don’t be afraid. Take risks because if you find something that’s exciting, then you’ll put in the extra work and make it a success. It’s amazing to see how somebody who is excited about something can get it done even if it seems impossible. So, go after things and don’t be afraid.

The second piece of advice I’d give is don’t let anybody tell you that you can’t do something. I think the minute you start believing them, then you’ll start to doubt yourself and you won’t let yourself succeed. People may tell you “You’re not the right person for this,” or “That’s not the right direction,” or “It doesn’t seem like you have the skillset for it.” But, as I said, if you’re excited and willing to put in the extra work, you can make things happen that others don’t expect.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s