Bhat's Notes

Dr. Vasudeva Varma, Dean (Research) – IIIT Hyderabad

Dr. Chakravarthy Bhagvati, AI Lab, University of Hyderabad

Dr. Jayanthi Sivaswamy, Dean (Academics) – IIIT Hyderabad

Invisible money in  academic career is very big. There are ample opportunities in small and big companies that want to work with us. We hike our fee to say no to solicitations from industry and then we have industry matching it. We are having now difficulty to say no as they are ready to accept steep hikes in fees.

Success is about having options. If you can choose what you want to do, then you are successful. If not, you feel misery. Research career gives you that option. Many would like to get back to academics after 10 years in industry. Doing research at that time may not be an option. If you want to be successful, research is what gives you an option. You can avoid QSQT (Quarter se quarter Tak) or the pressure of quarterly performance a big business goes through. Research means you have freedom and money to innovate at the company you work.

Research means you have no boss. No timings either. The funny thing is that you are judging all the time and no one is judging you. Don’t forget all the freedom, perks and vacations. Then you help hundreds of people fulfill their dreams. Some times bonding with them is much more stronger than what you have with your own family members.

Yes. We can’t participate in Olympics without doing local run. Research culture is required to train you well. Jason of University of Maryland said “small bones to big bones”. Start reading research problems and try to re-implement. You may look at a small variation, not publishable. Next you look at two three approaches and try the approach with hybrid one. These are like doing a mix of fitness workouts to build your stamina in research gym. Then you see where you have fertile lands for research. Interdisciplinary research areas are the most fertile ones. Who knows? They might be named after you.

You have to practice. Those who know the classical music know the importance of Sadhana or practice. That arohan and avarohan has to be practiced every day. So you have to practice like that in research. Understand that there are missing fundamentals. Improve your stamina incrementally.

Have the habit of writing every day. Keep a research general and you call it master general. Papers, working notes, lab presentations, conference thesis and final papers are by-products of what you write there. If you have nothing to write that day, go rewrite and find a new way.

Many of us end answering that question as someone else chose it for you.

Three golden rules for successful scientific research.

Rule 1) Raise your quality standards as high as you can live with. Avoid wasting your time on routine problems and always work as closely as possible at the boundary of your ability. Do this because this is the only way of discovering how you can move those boundaries forward.

Rule 2) This deals with the relation between the scientific world and the real world.

We all like our work to be socially relevant and scientifically very sound. If you can find a topic that satisfies both these topics, you are lucky. If you find them in conflict, prefer the scientific one. Imperfect solutions do more damage than good. If you offer a shaky solution to an urgent problem, it won’t be helpful. Going the path of social good may not be the right thing as the society is looking for. You are setting a wrong expectation as the society looks for you to solve as a researcher.

Rule 3) Never tackle a problem that you know others who are as competent and as capable as you can solve. Leave it to them. There are number of problems in the world. There is always a problem waiting for your  own unique abilities. Open the eyes and the problem will find you.

There are no time limits. Rules go for a toss. I work when I want. I never stop working. The fun thing is that I like it. I work because I want to. This is the nice thing about it. I work at night, at home and I don’t usually work at university. Your children start wondering when you also work as they work. You annoy them. My daughter says “this problem is easy for you, but not for me”. Then I think about it and we know why that shot of Federer’s backhand is very easily smooth. Noone sees the hard work behind. Easy is the outcome of hard work. This is a necessary condition of being a researcher.

Wherever I take my other daughter, someone there recognizes me. She gets a preferential treatment. Suddenly, your stock raises in front of your kids. Some times she likes it and some times she doesn’t. Teenager thinks you know nothing.

Train yourself: Research requires lot of training in terms of basics. Take the example of Newton. He didn’t know geometry. He goes back all the way to Euclid and after mastering it an year later he invents calculus. Know the basics.

Train your mind: Train your mind to consistently think of problems. When I drive on the road, I notice the number plate and keep finding patterns in them. Luckily I come from the pattern recognition side.

Read, Read and Read: Read anything and everything. Just keep reading. It helps in two ways. It widens your horizon. It gives many different perspectives. It improves your writing. Writing is an essential skill for researcher. You have to communicate. Keep reading. You will find that there are problems there too. There are interesting solutions for them on this side. For the link.

Bring ideas from different places: Most of you have done convolution in engineering. It requires you to take it from there into a network. In engineering, you give the kernel. In network, the kernel is learnt. That helped a lot as you can see. Go little beyond your text books. Bring ideas from different place and they work. Prepare your mind for it. Read a lot and write a lot. Discuss a lot. Play a lot.

I went in PH.D. as I wanted to do video games. Video games delayed my PH.D. by an year. Quake and Tetaris games killed together an year. You get to do things that you like. You get paid for it.

There is one person who was interested in architecture. She studied architecture and she loves to play with Legos. She was hired as one of the top researchers to build the Lego systems at Lego. There is fun in being a researcher. By definition, researcher is an unusual animal. Only half a percent of the world are in research.Dr. Jayanthi Sivaswamy

I used to be a voracious reader. The book I read scared me about research. It is by Ian Rand. It builds up to a point. There was no arxiv at that time. Then the character in that book gets crushed as someone else has come up with the same research problem ahead of him. It is a race. When a student came crying to me that someone else has just published the same problem, I told him “congratulations. You thought independently. Now focus yourself on how many ways you can differentiate your paper so you can sell it.” Approch a problem with the thought of  how you can differentiate yourself and tackle the problem. If you think research is about competitiveness ONLY, then you are setting yourself for a huge disappointment. Millions are out there attacking the same problems, even B.Tech. students are writing papers!

This competitive style of research means you are thinking about research in incremental methods. This can get you to a good start. It won’t lead you to a long sustained career. Why? You are allowing someone else to choose the benchmark. May the problem has to be looked at differently and defined differently.

I set my goals incrementally. I wasn’t ready for industry. So I did masters. While I was doing masters, I discovered my true love which is teaching. I discoveres the only policy that can buy me the life long fun is Ph.D. I was talking to an experienced person at 82. He says he never feels old. Students keep asking me questions that weren’t asked as they keep me young. I hot to think on my feet and come up with answered. My peers who retired long time ago are not young in their thinking.

Now researchers are paid big bucks too. I am here to sell not a research career, but a research career in academic setting. Choose academic setting to practice it. Then India will be a different country.

Go with the flow. Go with your passion area. Go with the flow where there are lot of supporting systems and there is lot of demand. It is also about what excites you. When I started, I started in electrical engineering department. I worked with someone in computer vision. I also had interest in biological visual perception. I decided to combine them. To this day, I retain interest in how natural systems and how computer vision systems work. Most of early stages of neural systems claim mimic  human visual systems. Not one to one but inspiration.

I personally found that spider got eight eyes and I have only two and some creatures have only one. We coexist in this world. A frog see a only when it is moving. Whereas we always see. There are multiple ways to solve a problem. We can navigate the world without an eye like those creatures do without an eye. Or we use many cameras.

Key ingredient is “we never know where the inspiration comes from. To be a good researcher, you need to learn the art of abstraction. You will use literature from out of the box and you get those key words. You never know when the inspiration will hit you.

You can also learn from how others deliver their ideas. One of my students asked me “what is this watershed algorithm? Where is watershed? Why do we use metaphors?” We as humans we like to anchor our ideas on something. Give codes. It is one way. Give pseudo code. It is another way. Instead, give the concept. When you say “this is how swarm of bees do”, that is particle swam algorithm. The maturity of a researcher is conveyed in how he conveys it. Within the first five minutes, anyone should understand.

Vasu: Understand that when you are done with Ph.D. and give a job interview, in fact that you are interviewing them. Avoid if you are going to be the best on that team. Look for who can be your role model.

Chakravarthy: Owning the work. Getting the sense of ownership so you become a researcher.

Q) When are you changing the path? Temptations to change the problem or even research guide.

Vasu: Surface might change. Underneath may remain the same. Solving a specific problem is not the important thing. Core may not change but the way it fits your bill might.

Vasu: This is the same question I had while  getting married. You will discover later that any person come as a different person depending on her mood. That is the same with being married with the same problem. You enjoy the motion.

Mark Granovetter (we underestimate the term weak tie). Information flow happened better through weak ties.