cat /proc/claus

So I’m leaving Japan.

After 6 years, I have finished my master and PhD at the University of Tokyo - in the last month, I was busy trying to find a place to work as a researcher, and generally preparing things for my trip. I will (probably) take a position as a researcher in Rio de Janeiro’s Federal University, in the COPPE engineering research institute. My new research is on using evolutionary computation to classify medical data obtained from a citometer — an applied theme that I hope I can talk about with more pride of “doing something useful” than my previous portfolio optimization research.

I say probably, though, because my status as a researcher is not yet confirmed. The supervising professor already said he wants me in his team to start as soon as possible, but the government research agency is dragging its feet to accept my documentation, and putting all sorts of bureaucratic roadblocks in my way. This is a taste of Brazilian’s nightmarish bureaucracy, and I wonder if I will be able to re-adapt to it, after years of the stupid, but very efficient, Japanese one.

There is a lot of stuff I’m gonna miss from Japan (or should I say Tokyo?), among them the easy access to any and all of the hobbies that I might want to have, the incredible convenience of life, and the absolute lack of need for a car. But I’m also looking forward a few things in Brazil — meeting my old university buddies, a more integrated and exciting research environment, etc. At least the changes in lifestyle will probably generate a lot of fodder for blog postings.

My short term plans are to go to the airport today, then spend a week at my mother’s house, then a week at my father’s house, before starting my research activities at Rio (although my activities have already sorta started with Matlab studies). I already have my place sorted out (I’m sharing an apartment with a friend), and I got some caches and people to meet in Rio (friends of my dance professor in Tokyo). Things I still need to sort out is how to get my RPG dosage in my new city, but a plan to convert my younger brother to the Hobby is already in motion.

I also got the best farewell present ever from Chaminda — and IEEE plush bear! Stay tuned from the pictures and more news when I get to the other side of the pond!

I was talking to a friend these days about our childhoods, and I heard something that stuck to my mind.

There is a holiday in Japan for the birthday of the emperor. Currently, this day falls on December 23rd. However, during the previous emperor’s reign, of course, the holiday was in a different day. After that emperor died, however, that date remained as a holiday, and it’s name was changed to “Green Day” or “Nature’s day” (and if you never knew where “midori no hi” came from, now you know).

Now, when she was a kid, this friend thought that, since the former emperor’s birthday was still a holiday, eventually, after a number of emperors had passed, most of the year would be made of holidays! So, in her kid mind, she wanted the emperors to die as quickly as possible, so she could have more days off school.

I found this pretty amusing

They say that for every problem there is a solution which is simple, elegant, and wrong. I found mine today while working on ETD, and I also find the correct solution (which is also elegant, although a bit less simple).

One of the rules of the “Tower Defense” games is that you cannot completely lock the path that the enemies take from the entrance to the exit of your maze. This path may be as long as it takes, but it must exist. In my implementation, I’m writing in each cell of the path the cost to go to the exit, so that each enemy does not need to calculate their exit paths all the time. Then I have to have a “locking detection” routine that detects if placing an aditional tower in a given location will lock down the maze or not, to prevent that tower from being put.

My first idea was to see what are the distance values of the cells the new tower is going to occupy. If those values were not repeated anywhere else in the maze, that means that that path is a bottleneck, and cannot be blocked. Thinking this solution for a bit while showed that it was wrong (exercise to the reader :-P). The new solution is thus: The entrance and the exit of the maze bissect the surrounding walls into two groups. I give each of this group a flag, and I check each tower when it is put down to see if it connects to one of the groups. If it does, I give it the same flag as the group. If a tower connects BOTH groups, then it means that it blocks the path from the entrance to the exit. I only have to deal with the special case of “island” towers, by giving them a third, “neutral”, group flag, which is painted 1 or 2 when one of the groups reaches the island.

I have not finished implementing the above, but I have implemented most everything else that I needed from the map. I did not have as much time to program today as I had yesterday because I used the bright sunny day to walk around a little bit, clean my home (including washing the dreaded bath), and meeting my friend Dionisio. All in all, I had about 4 hours to hack away at ETD today, and I think I managed quite a bit in this time. Tomorrow I’ll finish implementing the maps, then a bit of tower logic, then link the genome with the tower and the map I hope!

I woke up 2 in the afternoon today. Even when I was staying awake for days at a time when I was writing that article last month, I never had crossed the AM/PM line for waking up. I wonder if it is the cold - today was a constant 8 degrees during the whole day, and everyone I called for a coffee out said they couldn’t go, so I end up staying the whole day at home…

Which means that I had plenty of time to keep hacking away at my Python problem :-). Today, I tried a more direct approach by try and coding directly the functions I needed, stopping to google a concept or another when I ran into something I did not understand. This worked surprisingly well, as I managed to intuitively use the list constructs and function in python to easily implement my genome, mutation, crossover operations, as well as population mechanics, like elite, sorting, tournament selection, etc. I did find my share of bizarre bugs, like once when I got confused about instance and class scopes, and that resulted in a constructor operator which generated bigger and bigger individuals in geometric progression and ate up all my computer’s memory in just 4 generations, but by the end of about 6 hours I managed to have a fully fledged evolutionary system (although with a dummy evaluation function). Tomorrow I’ll try writing the engine for my ETD game/evaluation function.

Besides that, I also read up two chapters in the new book I have started “Here Comes Everybody”, by Clay Shirky. The book talks about and tries to explain the phenomenon of the massive, loosely linked community actions, like wikipedia or flickr, based on the idea that the cost to maintaining social connections has collapsed in the past few years, which allowed non-profit actions which were too expensive for informal communities to organize, but too unprofitable for formal companies to tackle, to flourish. Reading the book I can’t help but feel that I had heard all this talk in many different blogs, forum/slashdot comments, and Free Software talks, but it is always nice to see everything put together in one cohesive, well argued text, and with plenty of interesting anecdotes to illustrate the concepts.

Talking about books, last week I also read “A Wizard of Earthsea”, by Ursula K. Le Guin, and I really really recommend this book. I devoured it in less than 2 days. This book is one of the precursors of Medieval Fantasy, and the wizards and how the magic works in Le Guin’s world is too charming. The concept that a Mage is just as powerful as he knows and understand the world around him draws you into her world. I hope I can make my own D&D world as mystical and still consistent as hers.

And that’s for a very cold and gray Sunday. I got one of my three bases covered

Recently I have become much more familiar with Tokyo’s electronic neighborhood Akihabara. A few months ago I would have problems finding the Star Kebab I liked so much (a small turkish shop for kebab based sandwiches), and I would have to run around the place three or four times whenever I wanted to find some used game or particular computer piece. But now I can orient myself much better around there.

This is probably because on the past week I have been to Akiba 3 or 4 times - each different guest from Hospitality Club of course wanted to go there, and since I can get there by bike, I would usually follow them and show them around. Then I went a few extra times when Anna and Bonnie were here - Bonnie wanted to hunt for junk shops to try and assemble new ideas for her design products, which I found really cool, and Anna had a broken notebook which took many visits to different shops before we could finally fix it.

Today I was there again, this time to buy this Wireless Headphone I had seen a few days ago for a very cheap price. This dodgy shop in a back street had this stall with 10 or 20 units, and I thought I could come the next day to pick one up. Turns out that if I see something really cheap in Akiba I should buy it right away, because when I went to check it today it was already gone. But while around, I found this cool model shop where they had a rather big section on Ghibli models - and they were pretty cool ones, large mechanical clocks with the characters of Hayao Miyazaki’s movies doing different things. The shop was in the underpass behind the Glass shop near the JR Denkigai entrance.

And then we get closer to the post-a-day goal