Six months at the 国立情報学研究所

About a year ago I came back to Germany after spending one year in Japan. Now I'm going again — for six months. :)

Last time I went as an exchange student. This time I got into an internship program for Ph.D. and Master students at the National Institute of Informatics. — i.e. this time I'm actually doing something related to my studies (my exchange university last time had no computer science courses).

Another major difference is that this time I had to organize accommodation by myself. Find a place to live in Tokyo — what a nice ex ante adventure, yay ᕕ( ᐛ )ᕗ. What I knew from the beginning is that I would search for a room in a share house. I had previously made good experiences with share houses in Japan, knew that it would be cheaper, save me the hassle of having to buy a gazillion every day life items for just half a year and get rid of them afterwards, and knowing that I might end up speaking primarily English at work, a share house with Japanese people would up my chances of having extended conversations in Japanese on a daily basis.
So, I started searching on and, took a brief glimpse at non-shared flats on グッドマンスリー only to be scared away by the prices, and finally found something at ひつじ不動産. Given what I wanted — something not too far from work, not too expensive, where I could let someone stay at my place for a night or two, where thery're okay with someone who's guaranteed to leave after 6 months and on top of all that where a foreigner is able to sign a contract without a Japanese guarantor — I'm super happy with the place I found. :)
Additionally the countless mails back and forth until I found the place, the deal was sealed and all the details were discussed were quite the 敬語 boot camp. orz

Anyway, preparations are as good as done. Journey starts Oct 17th. Updates will be tagged #NII17. :)

Anki for exams

The computer science exams I take usually heavily rely on understanding. Learning for those exams means doing tasks that require applying the knowledge acquired. Memorizing facts almost never plays a big role. Quite a few exams are even "open book" — bring as much non digital material as you want.

Last semester I took two psychology classes with exams that were more or less the opposite. Being used to memorize stuff (kanji) with Anki, I thought it might be most efficient to use it for exam preparation. Because I was curious how much time I'd spend I recorded the whole process. Here's how to pass a 5th semester psychology class (Occupational and Organizational Psychology) as a CS major:

The challenge
The lecture consists of 386 slides. The exam can be passed with a very good grade solely based on the information given on the slides plus what is being said during the lecture, but students are advised to also look into referenced books and papers (I didn't). The exam consists of (iirc) 50 multiple choice questions, each of which has 5 answers of which between 0 and 5 are correct. Choosing a correct answer or not choosing a wrong answer results in 1 point. Not choosing a correct answer or choosing a wrong answer results in -1 point. The overall points for one question are reset to 0 in case they're negative. Example questions given for the exams go as deep as asking for names of authors or effect sizes of specific studies.

The solution

  • Attend all the lectures, annotate the slides (yellow boxes in the image above)
  • Start preparing for the exam about 3 weeks in advance, within these 3 weeks:
  • Spend a total of 14 hours to write a summary of the lecture's content (25 pages)
  • Spend a total of 9 hours to create Anki cards (477 cards)
  • Spend a total of 12 hours reviewing these Anki cards (2775 reviews)
  • Squeeze in 5 days of preparation for another exam with less material to learn
  • Pass with a not so great result, having spent a total of 35 hours on exam preparation (not counting lecture attendance)

Anki stats


28 hour day

I recently stumbled upon xkcd's comic on the idea of a 28-Hour Day. Having been interested in polyphasic sleep for a long time (without ever trying) and this being similar, I checked whether or not it would work out with my current situation of temporally fixed obligations throughout the week. Turned out that there would be very little overlap (blue=sleep, green=obligation). And the two cases where a lecture falls at the very end of a day ... just schedule two/three hours of sleep before that, right?

Wrong. Instead of extending my day by 4 hours per sleep/wake cycle I ended up with 3.5, 0.5, 2.5 ... -6. Friday, 1:30 a.m. I was super tired after having been awake for a mere 7 hours, went to bed and slept for 8 hours straight. This is how the whole week turned out. More like a rubber band that I stretched a bit than the intended wrapping around over the course of a week.

Lesson learned: if you sleep a lot anyways (8-9 hours per night in my case), the 28 hour day may not be for you. ^_^



28  27  26  25  24  23  22  21  20  19  18  17  16  15  14  13  12  11  10  9  8  7  6  5  4  3  2  1