The Machine Learning Center at Georgia Tech’s Seminar Series draws leading researchers from across academia and industry each semester to present on current topics in machine learning and artificial intelligence, applications for the technologies, and related insights and experiences. The popular series averages more than 100 attendees for each talk. This fall, five experts will giving talks across a variety of topics.
Galen Reeves, an assistant professor at Duke University, kicked off the series with a talk on information theory and statistics.
While Reeves was on campus, ML@GT had the opportunity to talk with him about his research, thoughts on working and living abroad, and lessons he has learned along the way.
Tell us about yourself and your research interests.
My work is a combination of electrical engineering and statistics where a lot of it comes from information theory. Information theory is the theoretical side of electrons and how do you efficiently represent, manipulate, and communicate information. A lot of the work that I have done is using ideas from information theory to think about high dimensional influence problems.
What first drew you to a career in tech?
When I went to college, I was interested in studying how the human brain works. I started by taking all of the required biology and other science courses, but realized I wasn’t feeling very fulfilled. I started taking some artificial intelligence courses, and they were neat but I still wasn’t feeling fulfilled and like I was understanding what intelligence was.
I was actually in the electrical engineering program at Cornell University and eventually took a single systems and information theory course. I loved it! It showed me where you could use math to come up with simple and elegant characterizations of complex systems which I found to be beautiful.
At this point, I realized that this wasn’t going to help me understand the brain, but it was cool. So I switched gears and ever since then I have been studying and using a wide variety of tools like physics, statistics, and machine learning to understand complex systems.
Tell us about some of the projects that you are currently working on.
The project that I’ll be talking about today gets me excited because it has a fun application. It’s a network inference problem known as community detection. The idea is that you observe a relational network like a group of friends. Based on those friendship connection patterns, you look for underlying communities like “Are they all at the same college, or graduate students, or in the same lab?” It makes sense that people in these communities would likely be friends with each other. This is studied frequently in the social sciences field.
My colleagues and I are trying to look at it from more of a computational and statistical perspective. Broadly, our goal is “Can we do this precise theoretical analysis for more interesting networks?”
[For more information on this project, check out Galen’s talk]
What advice would you give to someone going into this field?
I hesitate to the use the word advice because there are so many ways to be successful. The best thing I can say is to like what you do.
What is a lesson you have learned from being a lifelong researcher and student?
I know for certainty that there are going to be some really cool innovations over the next couple of years. We might look back on it and think “How did we not think of that a few years ago?” And sometimes the ideas aren’t even new, they just happen at the right place, the right time, and are marketed the right way where they catch on. It’s cool because whether or not you realize it in the moment, your work really matters, even if it doesn’t feel big or incremental. As academics, the idea is that as a community there will be some cool advances you can look back on and that’s pretty neat.
You have studied and worked all over the world. What has been the best and hardest part of all of these transitions and living abroad?
I love this part of being in academia! It never ceases to amaze me that wherever I go in the world I find friends or people that I know. Academia is very welcoming and it’s exciting to connect with people in new places who have similar research interests as me.
What is something that you wish you knew at the beginning of your career?
It’s really hard to predict what is going to be successful. Some people try to hitch their research to what’s ‘hot’ or what will be popular, even if it’s not something they are interested in or good at.
There are a lot of times where you work on a project that you’re proud of and think is great, but maybe it’s only good for a small research area, or it’s ready at the wrong place or time. Other times, you work on something that you are not as proud of and it blows up and people think it’s really cool.
To summarize, I’d say don’t not do something because you do not think it’s 100 percent the most awesome thing right now because it might be surprisingly interesting to other people.
Why do you think it’s important for people to come and give talks like the one you gave today?
To me, it’s important in two ways. This is my first time at Georgia Tech and I’ve already met so many people and learned about what they’re working on. It’s exciting to connect with everyone and learn about what they are doing. There are so many papers published, it’s hard to keep up with who is doing what!
I remember coming to talks like these at the beginning of my graduate school career and feeling so lost. They can be daunting because you’re sitting there listening to the presentation and students around you are nodding their heads and asking clever questions, and you have no clue what is happening. I encourage every student to keep attending these talks, even when they don’t understand, because eventually the pieces come together and you learn a lot about what’s going on in a variety of areas. As a senior graduate student, I looked forward to these talks and they became a rewarding part of my day.
Is there a paper that has had an impact on your work?
Absolutely, there have been so many. One in particular is a paper from 2005 called “Randomly spread CDMA: asymptotics via statistical physics” by Dongning Guo and Sergio Verdu. It introduced a whole new world to me in graduate school that has now become pretty active.
Who is someone that inspires you and why?
My post-doc advisor, David Donoho at Stanford University. A lot of what I learned as a graduate student was just by reading through his papers line by line and figuring out every tiny detail. It was crazy that four or five years later, I was knocking on his door as one of his post-doc students. He’s brilliant and it’s amazing the way he can see both the math and the bigger picture as well.
Tell us about something that brings you joy.
While I do find my research very exciting and joyful, there are few things that beat the feeling of when you deliver a lecture that really seemed to resonate. When you feel like you did a good job explaining a concept and students are engaged, it’s a certain type of joy, but it’s great.