My first year of graduate studies at MIT

Wow its already 3 months since my second year at MIT started – time is flying (originally when I started writing the article I had 3 weeks here). There are so many things to do, to research and to explore that one hardly realizes how time is passing by. My first year at MIT was an incredible experience – in terms of heights as well as lows. I have never been so excited and never been so challenged before. At first I describe some lows and highs before I give a list of takeaways from this year.

The most challenging situation definitely was, when I was sitting in the final exam of 6.437 (at MIT everyone calls classes only by their numbers – 6.437 is an Inference and Information class) and had to realize, that I had no clue how to solve any of the three problems of the exam. After the first hour (out of three), I started to get worried. After 2h without anything written down I had to fight panicking for the first time (for my TUM colleagues: EMF 2 was really bad too, but I had at least written down something after the first third of the exam…). But then gradually I made some progress – which saved me from failing the class I assume. I think, the reason for this was that I was trained to finish mildly hard exams really fast. At TUM, if you ever stop writing in an exam you are screwed. In 6.437 (and the other classes with exams at MIT) the challenge was not the timing but the hardness of the problems: In 6.437, they were all introducing challenging novel concepts, that you then had to utilize to solve the exam problems.

Similarly challenging is the theory-heavy research I started with. Turns out, that Bayesian non-parametrics sounds cool, but really is a ton of statistics and probability theory. I am now often spending an afternoon with pen and pencil instead of programming in front of my computer. While this was weird at first, since I felt like I did not make progress – it is actually the time in which I make most theoretical progress (which luckily is what gets rewarded most in academia). After a while I now start to like it more and more, since it is actually less frustrating than wrestling with a ton of libraries to get new algorithms running on my (awesome) Turtlebots.

The most thrilling, exciting and intense time definitely was the last month, when I was for the first time solely focusing on writing a paper. About a month before the deadline of the conference I submitted to, I had an interesting idea (which I cannot tell yet due to a double blind review process). So I went for it. In the process of refining the idea, it became more and more apparent, that we were onto something new, useful and potentially very powerful – unchartered territory, a white area on the map of computer vision research. Okay I am exaggerating, but I was thrilled. At night I could not fall asleep, since new ideas and solutions came to my mind, that I had to write down to not loose them. In the mornings I woke up already again thinking about the paper. I bought new underwear, since I ran out of it and did not want to take the time to wash. The last night before submission, I did not sleep to get the related work section straight (I hate all-nighters – I can count all that I have pulled on two hands. And most of them were actually partying). In the end I submitted the paper in pretty good shape – I am excited for the feedback of the reviewers in January.

In the beginning of 2013, I was elected Vice President for Social and Orientation of the EECS Graduate Student Association. In this position I had a lot of rewarding experiences and learned a lot about the inner workings of the EECS department at MIT. My team and I organized the visit day and the orientation for the newly admitted EECS students. I was focusing heavily on making sure, that the new students had every opportunity to network and stay in contact from the visit day onwards. It makes me very happy now to see them working together in big groups on class assignments as well as to see group pictures of them together on Facebook. On the social side, my team and I had a lot of fun organizing ice skating, sailing, Pi day, movie nights, Oktoberfest, mixers with other departments as well as established a coffee hour for all EECS PhD students that is regularly frequented by 40 to 50 people (a large number for EECS standards). I find it very rewarding to see these events come together, to work with a motivated team and to see people enjoy the events.

My main takeaways

Research:

• the word “research” makes a lot of sense: you have an idea, you search whether someone has done that before and then you re-search since it has almost surely been done before…
• solving problems is easy finding the problem is hard (and frustrating).
• theory can be really enjoyable when the other option is to get a visual SLAM (simultaneous localization and mapping) system to run.
• nothing is more fulfilling, exciting and intense than pursuing your own idea (on the work-side of my life)
• working for the same deadline as a labmate makes for a great bonding experience
• running out of coffee two days before the deadline is not good
• always have some protein bars or a nut mix in your drawer
• make it easy to buffer intermediate results of long computations to speed up plotting and evaluation of the algorithm
• post-docs are an awesome resource when it comes to paper writing, code snippets or learning the theory they have specialized in.
• write everything down – humans are no Turing machines due to limited memory. Therefore, we should always be writing stuff down onto an external tape.
• 4 to 5h of quiet time for reflection at the end of a week is very important to integrate new knowledge, collect ideas and plan for the next week
• Python is beautiful and so far the best tool for CS research I have found – get to know numpy, scipy, theano, matplotlib and mayavi and forget matlab.
• put everything into GIT and commit often – sharing between several computers as well as collaboration with others made easy.

Classes:

• do not mess with MIT math undergraduates.
• grades are not important anymore during the PhD – the minimum required is good. Research is most important now (later nobody will ask for your grades but for the projects you worked on)

Life:

• a wonderful understanding girlfriend, awesome friends and roommates are important to keep you sane and to cheer you up in challenging times with
• a great team anything is possible
• always aim higher than you think you can reach – the most surprising things might happen
• honorary work is fun and rewarding – interesting opportunities come up
• that you otherwise would not have heard of
• never neglect your own health (exercise as well as healthy food) – it will come back at you like a boomerang
• celebrate success with the people that matter to you
• take time to read awesome, non-technical books purely for entertainment
• taking a nap in the sun after lunch every once in a while makes happy
• sleeping is better than procrastinating