Mathias Bank

Yeah! Yesterday I have finished an exam about data mining and reinforcement learning. This was the reason, why there were nearly no posts at this page. But now, I have time. Time to do the jobs, which has pent-up.

If you will ask a politician, why he or she thinks, that tuition fees are necessary, they will say to you: “Of course it’s necessary, because the universities need money. And for students, it is no problem to get enough money. You can see, that it works in other countries like Great Britain.”

Ok, let’s see:

[...] In a survey that asked 130 students whether they knew any friends involved in the sex industry, one in 10 said they knew of students who had stripped, lapdanced or worked at massage parlours and escort agencies to support themselves. Over 6% said they knew students who worked as prostitutes [...]

A new sample of politicians who are saying: “Sorry, I have no idea!”

These days, I’m studying artificial intelligence again (especially muliagents). For this, I want to learn LISP, a really nice programming language (wich can also be used as language for web pages). One main element of LISP are lists (LISP = List Processing Language). To understand, what we can do with lists, I’ve found this nice dialog:

Spoon Boy: Do not try and bend the list. That’s impossible. Instead . . .
only try to realize the truth.
Neo: What truth?
Spoon Boy: There is no list.
Neo: There is no list?
Spoon Boy: Then you’ll see that it is not the list that bends; it is only
yourself.

Spoon Boy wants to tell you, that you can do nearly everything with lists (ok, I think that is what he wants to tell)

For a practicle work, I should use torch3. But unfortunatelly, the support is really very bad. There is a forum and there are adresses to which you can mail, but noone wants to answer. So I’m looking for another classification library.

Very interesting is LIBSVM, which is avaiable in several languages (Java, C++,…). In some newsgroups you can even read that LIBSVM is sometimes faster as torchsvm (nearly 10 percent). I now try to use this library.

At first, I wanted to use normal neural networks. But in a lot of papers, I can read that Support Vector Machines are much faster and better in pattern recognition. Let’s try it.

These days, i’m learning for my exams in artificial intelligence. Because I haven’t understand constrained satisfaction problems (a speciel search problem in ai), I looked at wikipedia. There was not much help. But therefore, there was a link to another CSP: The Combat Simulator Project. It is an open source flight simulator. I was interested, so I searched a little bit. There were to other interesting links:

• True Combat: A full ego shooter
• StarWars: free starWars games

Ok, back to artificial intelligence: Of course, there is a good information source: Wolfgang Runte has written a good degree dissertation: “YACS: Ein hybrides Framework für Constraint-Solver zur Unterstützung wissensbasierter Konfigurierung“. Specially chapter 5 helped me: “Klassische Constraint Satisfaction Probleme“.