Prof. Prasad Tetali is a Regents’ Professor in the School of Mathematics and the School of Computer Science at Georgia Institute of Technology. Dr. Tetali obtained his Ph.D. (1991) from the Courant Institute of Mathematical Sciences, NYU, after earning an M.S. (1987) from the School of Automation at IISc. His reseach interests lie in probability, discrete mathematics, algorithms and optimization, and has published more than 110 research articles. He is recognized as a SIAM fellow (2009) and an AMS Fellow (2012). Dr. Tetali is a former director and a current member of the steering committee of Georgia Tech’s Algorithms and Randomness Center Think Tank (ARC) and has been on the coordinating committee of Georgia Tech’s renowned interdisciplinary Ph.D program in Algorithms, Combinatorics and Optimization (ACO) for the past 2 decades. He served as the interim Chair of the School of Mathematics at Georgia Tech during CY 2015-16. He is currently the director of the ACO program and is an active member of the NSF-funded Transdisciplinary Research Instiutue for Advancing Data Science (TRIAD) at Georgia Tech.

Foreword: Prof. Prasad had his first research paper published with the most prolific mathematician of the 20th century, Paul Erdös, consequently earning him an Erdös number = 1. I personally find the above fact stirring as it is not quotidian to be collaborating with an eccentric virtuoso and child prodigy who published around 1500 mathematical papers in his lifetime (a figure that remains unsurpassed). Read on as Prof. Prasad shares his experience of working with eminent scientists and his take on questions that plague the next generation of budding researchers, such as:

• Will Theoretical Computer Science research render you jobless?
• Is research even meant for you? If yes, academic or industrial?
• What is the fundamental problem solving skill for every researcher? and many more… Stanly Samuel, Ph.D. student, CSA, IISc

This interview is conducted between CSA Writing Team and Prof. Prasad.

You did your MS at IISc in 1987; how does it feel to be back after so many years?

It feels great! I must say I feel honoured, especially to give the institute-wide lecture in the IISc main building. I was telling some of them who attended (the lecture) that the only reason to go to the main building, back in the day, was to make an STD call at the corner. There were no phones in the hostel rooms at the time. There was one phone in each hostel at which your parents could reach you but you could not call outside campus.

How, according to you, would you say that the time you spent at IISc affected your current research interests?

My current research I’m not sure, but research definitely. After an undergraduate degree in ECE, I came to the School of Automation. I joined the M.S. by thesis programme that was very much geared towards going into research. I think it prepared me very well in terms of starting research when I went to NYU for PhD, so that I had papers already within the first two, three years.

Were you into research before coming to IISc or did IISc change your mind?

I don’t think I had explicitly thought of research before IISc. We all had a final year project and my project was on an intel 8084 or one of those micro-processors. I believe, I did some assembly code to simulate floating-point arithmetic, which was my first research experience. My father was a university professor of Geology at Andhra University and that’s perhaps why I ended up in academia.

How did you get interested in discrete math?

I have to say, it is due to my MS mentor at CSA, Prof. Veni Madhavan. He introduced me to combinatorics, Knuth’s volumes, in particular Volume 3, and that’s how I got interested in discrete mathematics.

A lot of people think there is very little scope for research in Theoretical Computer Science. Aspiring researchers think that it is tough to find something new in this field. What are your opinions about this? What do you think are some interesting challenges lying for mathematical and theoretical research in computer science?

I digressed from your original question. It is true that there are fewer jobs out there to pursue TCS in the very classical sense of traditional algorithm design or understanding the complexity of NP-hardness problems. But let me say, for better or worse, algorithms are in use, literally, everywhere these days. So the TCS is community is constantly making forays into, and having an impact in, various classical as well as modern branches of science - whether it is algorithmic game theory, auctions and mechanism design with Economics, or bioinformatics and genomics in Biology, or Deep learning in Non-convex Optimization and in Operations Research, analysis of mixed-integer programs in Operations Research, Robotics and other areas.

The TCS community is transporting their body of knowledge and analytical techniques in explaining various natural phenomena as well as computational efficacy and intelligence of machines. If I had to speculate, ‘ML Theory’ will be a thriving area for a while, bringing TCS researchers together with those from ML and AI on the one hand, while also bringing in those with expertise from the social sciences and public policy, to tackle emerging aspects of ethics, fairness and privacy in the context of ML algorithms. I guess we have social media and virtual reality to thank for that.

You’ve had a very illustrious research career and have had the opportunity to work with a lot of eminent researchers. So do you have any interesting anecdotes or some interesting experiences?

My first paper was pretty much with Paul Erdös. Once my wife and I moved to Atlanta, he visited and stayed with us a couple of times. Atlanta is a hub. There’s a famous saying by the Economist magazine, I think, that whether you go to heaven or hell, you transfer in Atlanta. Erdös used to say that as he passed through Atlanta. Once he had come to visit us. He had arrived from Argentina or somewhere where it was winter at the time. And he was going to leave in 3-4 days. I saw that there were a bunch of things, like a jacket and winter coats, in the closet. He had not yet packed them. I asked him, “What about these Paul?” He said, “Oh I’m going to be in North America for the next few months. So, I don’t need them.” He travelled minimally and perhaps owned less. I really wish that I had known him earlier in his career. I knew him during 1989–1990 through 1995 or so and he passed away in 1996. He liked to think about math problems constantly. He would digress occasionally to talk about history or politics. During the time I had known him he was mostly applying his older known techniques to solve new problems. We had two different proofs of one of the bounds in our paper. The simpler proof used a new inequality (at that time) introduced by Svante Janson and others. Erdös has asked me multiple times to explain how that proof went. He was still trying to learn new stuff that intrigued him. I have been very fortunate to have had great mentoring and productive associations with famous, yet humble and sharing, mathematical souls! I have also been blessed with several great students. As I am aging in my career, I find that they do more and more of the bulk of the work, I am embarrassed to admit.

The one regret I have had for a long time was not having stayed an additional year to learn more math. I was at an excellent institution. I took some basic required things and some additional number theory because I liked it. Otherwise, I didn’t learn probability theory nor pure math in depth, for example, and I had to learn some of that later on my own.

One thing that a lot people who are doing Masters and PhD wonder about is whether they should eventually join academia or should they go for industrial research. How would you suggest they make that choice?

Yes, that’s a very good question. In general, it’s a very personal decision. Some people say that they don’t want to do this abstract stuff, they want to do something which will see some application or impact people’s lives immediately. I’d say ultimately, it is about what gives you satisfaction. What you find rewarding. That’s perhaps what should guide you. After my PhD, I did a postdoc at AT&T Bell Labs, which had many famous people who had been working there for many years. While I enjoyed that very much, I knew after my first year that fast-forwarding 20 years, I did not want to be at a place like that. To me it felt very insular. The lab was in a very comfortable, idyllic place in the middle of nowhere in New Jersey. You went in whenever you went in, and you came out in the evening or at night and you had no connection with the rest of the world the whole time. This is pretty much the everyday case. Maybe because my father was a professor and I grew up in a university area, in an academic environment, I knew I wanted to be at a university.

But now after 25 years at Georgia Tech, I would say, you should go for teaching/academia if you find teaching rewarding. It may not be enough if you find it tolerable. You should find it satisfying on some deeper level. Just to be clear, it is not that I am constantly craving to teach. I still prefer the research route as long as I can think and be creative or contribute to new developments. But I also like teaching, and I find it very satisfying. It is through teaching, I feel, that I can give something back to the community. Sadly, I’m not fundamentally changing the world through my research. It’s not like I’m involved in discovering a new planet, or something arguably more profound like decoding the human genome or cure for cancer, something that’s obviously of huge impact. Given that is the case for me, it’s the teaching and mentoring students which I feel justifies why they should pay me a good salary. I’ve also gone on leave to Microsoft Research several times. I like talking to industrial researchers. I don’t necessarily need to know what the applied problems are, but I’d like to know about some theory that they’d want us to do, inspired by a practical problem. Also, it gives me a break from the usual. I mean, teaching when done right can be exhausting. If you do a good job or if you struggle trying to do a good job, then you need a break from that.

Ironically, I am on my sabbatical semester without teaching load but then I give 10 lectures in all during one month here in Bangalore and I gave 8 hours of lectures in Shanghai two weeks ago; I am wondering how and why I agreed to speak so much… If one doesn’t enjoy teaching, and prefers to do research 90% of their time, then the students will be short changed. You don’t learn as effectively from such teachers and number two, such faculty also kind of gets grumpy and tend to take it out on the students, because they really don’t want to be there to help you learn or can’t spare such time. (I’d say their contribution in helping students is an important part of why we are in an institution of higher learning, but choosing to go into that is up to the individual, and they should be aware of this obligation, whether in writing or not. It’s nice if you realize it sooner than later, like with anything in life. This is why in graduate schools, in math departments, it’s nice that you are asked to TA, and almost teach (an undergraduate class) in fact, towards the last two years. It’s not uncommon. I have had students of both types, those that preferred academia and those that didn’t. Arindam Khan (who is hired here as a faculty) got teaching awards while at Georgia Tech as a TA. I have had another brilliant student Emma Cohen. She would have been very good if she chose to go into academia, but she didn’t feel compelled to teach, so she joined the Centre for Communications Research, located in Princeton, a Thinktank for mathematicians working under the umbrella of the Institute for Defense Analysis. So it is a very personal decision.

Finally, having taught for so long, is there any advice you would like to give Computer Science students in general (in terms of learning)?

One thing I will say, after teaching in Shanghai is that in Asia (India included) students are very passive. Maybe it has changed now but in the past it used to be that way. We feel it is rude to ask the instructor a question if you didn’t understand something. Most of us think that we can figure it out later. Whereas in the US, the system is different. Students don’t hesitate to ask questions. I think it is good to pay attention and ask the instructor questions before you get too far behind. I had to really encourage students repeatedly to ask questions and often they would ask really interesting questions, besides the basic ones.

Then the whole class (including the instructor, sometimes) benefits from such an interaction. So that’s my main advice. Also, as a nascent graduate student, you might think you are interested in topic X and may start narrowing down early on. But in your first year you should take different classes and keep an open mind. Don’t think that if you came to learn and work on topic X, you are now forced to do that. Even if you get assigned a professor A who is an expert on X, it is okay if you later say that you are not sure anymore. So take your time to decide. Other thing I find too often is that students who are working with one professor think that they should just follow that route, talk to that professor, and just stay in that lane. By the time I finished my PhD, I had written papers with 4 different people: Erdös, Fang Chung, Peter Winkler, and then, finally, I had a paper with my own advisor.

As long as your advisor doesn’t get too mad at you, you should be open to talking to or interacting with one or two other faculty. They will give you a richer background and exposure. Also, you would graduate with a stronger portfolio. If you say, “I have worked in Game Theory, but I also have one or two papers in this other topic”, that’s always better in the job market. Learn as much as you can during graduate study. I could not have done all of this in my four years of PhD if I had not worked on my foundation during my masters at IISc. During my generation, many who went for a Ph.D. right after a bachelors at an IIT or another strong undergraduate institution, typically took longer to learn to do research. They were not trained at a research minded institution like IISc. I had enough papers after 3 years in PhD that my advisor said that now I could do whatever I wanted. I could graduate in the fourth year, and I did, although it was a terrible time to graduate; the job market was really bad.

The one regret I have had for a long time was not having stayed an additional year to learn more math. I was at an excellent institution. I took some basic required things and some additional number theory because I liked it. Otherwise, I didn’t learn probability theory nor pure math in depth, for example, and I had to learn some of that later on my own.

Once you graduate, you start some job, academic or not, and you won’t have the same time to just sit and learn something deeply. So, while in graduate school, maximise your learning.

Other interesting questions answered in the longer version:

• You gave an excellent talk here on Markov chains, mixing time, and their applications. Could you motivate the same to undergraduate students?
• One thing that we found very interesting in your talk is Markov chain mixing for evolutionary dynamics. How did you realise that there was a scope of Markov chain simulation here?
• What do you think is an active area for applied research?
• Students who join for Ph.D. or Masters by Research are not quite sure what kind of topics they should be studying, what kind of courses they should be doing. So if someone wants to start a career in theoretical research, what kind of preliminaries, do you think, they will need?

Credits:

• Compiled and edited by: Mahak Pancholi and Stanly Samuel

• Interviewed and transcribed by: Haroon Ansari, Manohar Lal, Shivika Narang, Mahak Pancholi and Stanly Samuel

CSA Writing Team (CWT)