Kanada's Research Topics

Computation with randomized order

2007-03-31T00:00:57Z

Recently, computation methods with randomness are widely studied. Evolutionary computation including genetic algorithms and genetic programming are examples of such methods. Computation using simulated annealing, such as Bolzmann Machine, and methods for increasing efficiency using randomization in classical algorithms are also studied. The randomized computation models I previous studied, i.e., Chemical Computation Model (CCM) and randomized cellular automata (RACA), also used randomized computation.

]]> In these computation models, computation proceeds by successively applying given rules. The order of rule applications is randomized. Some interesting phenomena occur by regularly (synchronously) applying rules, but regular application often causes unexpected results. Randomness sometimes causes improvement of efficiency. However, I researched the computation methods above rather because, firstly I expected that we might be able to get intended results more easily without using specialized techniques, and secondly I believed that randomness is the source of emergence.

Pages on related researches:

Randomness and Computation (RC2005)

Software and hardware

2006-10-18T13:01:46Z

This entry focuses on my hobby rather than research.

To be described. ]]> このエントリーは研究というよりは趣味のはなしが中心です．

大学 2 年になるまで，私にはソフトウェアというものにはほとんど縁がありませんでした．一方で小学生のころからハードウェアには興味があり，ラジオやその他の電子工作をしていました．中学・高校のころにはオーディオファンになって，いまもある「無線と実験」，「ラジオ技術」などの雑誌をみて，ステレオアンプをつくりました．そういう私にとって，秋葉原で Intel 8080 が売られるようになると，それをつかってマイクロコンピュータをつくるというのは，むしろ自然なことでした．このマイコン製作についてはブログのエントリーに書いています．ここにも書いているように，デジタル回路の実装をよくしらなかったために，おそろしく複雑な配線をすることになり，もうハードウェアはやめようとおもうにいたりました．そしてソフトウェアの世界にはいってきたわけです．

就職してまもなく，ベクトル型のスーパーコンピュータ S-810 のコンパイラの仕事にたずさわりました．コンパイラはハードウェアとソフトウェアとの境界のプログラムであり，またとくにあたらしく開発されたコンピュータのためのコンパイラなので，ハードウェアをつよく意識しました．当時の汎用コンピュータのコンパイラ開発においてはハードウェアの特性をそれほど意識しなかったのではないかとおもうのですが (いまでは RISC のためのコンパイラ開発ではハードウェアをつよく意識するのは当然のことだとおもいますが)，スーパーコンピュータはハードウェアをうまくつかうかどうかで性能がおおきくかわってくるため，ハードウェアの設計者からもしばしば話をきいていました．

コンパイラ屋をやめてからはハードウェアを意識することもすくなくなりましたが， DOS/V パソコンの部品が秋葉原で売られるようになってからは，ふたたび自分でパソコンをくみたてています．自宅でつかう DOS/V パソコンは，家族のものもふくめて，すべて私がくみたてたものです．自宅でやるひとはたくさんいますが，私は会社でつかっているパソコンもたいてい，くみたてています．会社ではなかなか道具がそろわないので，机のなかにドライバーやラジオペンチをもっています．しかし，単なるくみたて以上のことは，ひさしくやっていません．近年は FPGA がひろくつかわれるようになって，半田ごてなしでハードウェアを開発することができるようになっているので，やってみたいとはおもっていますが，いまのところ，なかなか手がだせないでいます．もう，いい年になっているので，定年後のたのしみとしてとっておくということになるかもしれません．

Fusion of digital and analog

2006-10-18T12:58:28Z

One of my research focus that is common to my several research themes is, in short, "fusion of digital and analog". There was so-called analog computer, but now it has become old fashioned. Nowadays, the word "computer" certainly means a digital computer. However, human senses are inherently analog, so I think computers should come close to analog senses for the sake of ease of use. It means that combination or fusion of good points of digital and analog is required. Although analog computers were dead, neural networks that replaced analog computers (?!) had some good points of analog things. Since I was stimulated by them, I thought of fusion of digital and analog very frequently.

]]> In the research of CCM (Chemical Computation Model), I studied computation using production rules called reaction rules and evaluation functions called local order degrees. Here, reaction rules are digital, and local order degrees are analog. We can solve many problems, constraint satisfaction and optimization problems, using the combination of these two. In digital worlds, programs often do not stop at appropriate point but run away if writing a wrong program. However, combination with certain analog thing, i.e., using a function with a minimal value, probably makes the chance of such happenings less.

In axis-specified search, by specifying "search axis" in addition to search keywords, we can get search results that are well-organized along the axis. Here, a keyword search is digital, but it is combined with "axis", which is analog. For example, if we specify time as axis, we can get results sorted by time, i.e., we can get a chronological table. Or, if we specify geographical space as axis, we can get results on a map.

If we use voiscape (a virtual "sound room" based communication medium), two or more people can talk in a virtual sound space enabled by spatial audio technology. The objective of this research is to enable more natural human conversation through a medium by expressing feeling of direction and distance of voices by using spatial audio. In conventional media, such as telephone, the communication state was one of two digital states, i.e., connected to or disconnected from the other party. In voiscape, the other party can be in any distance and in any direction, i.e., it realizes analog states.

The N queens problem

2006-10-18T12:31:44Z

		Q
Q
			Q
	Q

It is time consuming to solve the N queens problem (See the description below), as same as other constraint-satisfaction problems. Therefore, many methods are deviced to solve it faster. In my research, I often used this problem as an example too.

]]> The first time I used this in Logic/Symbolic Vector Processing. By using a logic programming language such as Prolog, you can program exaustive search using backtracking. The objective of this research was to convert such programs to high-speed programs for supercomputers. However, the acceleration ratio of even a successfully converted program was constant, so the program still needs exponential computation time.

The second time I used this problem in CCM (Chemical-Computation Model). In CCM, randomness was inherently introduced, and we did not aim perfect computation (or complete method). In such a stochastic computation, it does not need exponential computation time. By using CCM, you can find a solution by a personal computer much faster than by a supercomputer using a backtracking method. You can try such computation by using the N Queens Problem and Sorting page, which contains CCM-based Java program of the N queens problem. You can try various methods by changing options on this page, but I will describe the detail of this issue in another opportunity. (But the above page contains most of such information.) To develop stochastic solving methods, neural networks and genetic algorithms can also be used. The method using CCM has similar nature to such methods.

Apart from Kanada's research, the N queens problem is also used in N. Wirth's book titled "Algorithms and Data Structures". R. W. Floyd's paper titled "Nondeterministic Algorithms" (Journal of the ACM, Vol. 14, No. 4 (1967), pp. 636-644) also used this problem. This paper was explained by Kanada in the 2003-February issue of "Information Processing" (Vol. 44, No. 2, p. 198). As I wrote in this explanation, the problem was also used as an example of quantum computation.

Explanation of the N queens problem

The eight queens problem is the problem to get layouts of eight queens on a chess board without taking each other. An extension of this problem to N-by-N board is the N queens problem. Problems to get a solution that satisfies given conditions, such as the N queens problem, are called constraint satisfaction problems (CSPs). Many CSPs belongs to a class of difficult problems, called NP (Non-Polynomical), which can only be solved by enumerating solutions. The N queens problem is one of them. The reason why this class is called non-polynomial is that problems in this class cannot be solved in time represented by N's polynomial expression (i.e., it takes exponential time to solve them).

[Related page: The N Queens Page]

Purpose of this blog

2006-10-16T14:09:02Z

When I became a research scientist, I decided my research themes according to my preference. However, I moved again and again to various research areas according to the needs of my company and societies. Those themes do not seem to be related each other at a glance. However, there are common backgrounds or philosophies among them. In this blog, I will write such a relatively heavy topics that spread among two or more research themes. Initially, I intended to write such issues in my general blog, but I changed my mind to separate them. Relatively light topics on one research theme are described in the general blog. I also made a Japanese version of this blog. By the way, the research themes in my company are described, but I write them as a research scientist and my company has no concern to these descriptions.

My lifework

2005-02-02T15:00:00Z

I have studied CCM (Chemical Computation Model) for emergent computation, axis-specified search, voiscape (a virtual "sound room" based communication medium), and so on. There are common "themes" among the above three researches. They are "fusion of digital and analog" and "fusion of symbols and subsymbols".

]]> CCM is a computation model that leads symbolic or "digital" computations represented by motions in discrete state spaces to a solution by using so-called local order degrees (a type of evaluation functions), which are "analog".

Axis-specified search is a method of search, which organizes the results of "digital" keyword searches by using time --a one-dimensional "analog" quantity -- (in the case of year-axis search) or by using area or map -- a two-dimensional "analog" quantity -- (in the case of geographical axis search). I tried to enable intuitive grasp of search results by putting them in analog spaces.

In voiscape, voice itself has both digital nature called language and analog nature called sound, but, in addition, by putting it in a three- (or two-) dimensional analog space, move it from two-valued (digital) world to a world with perspective, which can be well grasped by human senses.

Although the two research themes except voiscape, which is ongoing, cannot be called life works. However, I believe the common themes, i.e., "fusion of digital and analog" and "fusion of symbols and subsymbols" will be my life works. In addition, I changed my research themes as the above order, but I think I can say my research theme gradually becomes closer to "human beings". In my master's thesis, I chose "programming linguistics" as the research theme. Since then, except CCM, I can say I tried to revive human existence where it was conventionally ignored or abstract away.

Note: Because I included a content in an independent Web page into this blog, the submission date is before the date I started this blog.