While we are waiting - PacMan vs Ghosts

[Parasprites]

It doesn't seem to have any activity though.

Anyway, they definitely need a better visualizer.

[MagnumPU]

This is awesome, thank you for sharing. The starter package is pretty nice, beating it is going to be enough of a challenge for me.

[PeterV]

I found another interesting contest, also from the University of Essex:
the Physical Travelling Salesman Problem Competition http://www.ptsp-game.net/index.php
It's about writing a controller that finds an optimal way of visiting way points using a spaceship. It's based on the travelling salesman problem, but with some extra constraints on the spaceship movements

[aerique]

I found another interesting contest, also from the University of Essex:
the Physical Travelling Salesman Problem Competition http://www.ptsp-game.net/index.php
It's about writing a controller that finds an optimal way of visiting way points using a spaceship. It's based on the travelling salesman problem, but with some extra constraints on the spaceship movements

Which seems to be Java only as well

What's with academics and Java?

[Parasprites]

Well it's the same people, so it shouldn't be a surprised.

But really, would it kill them to allow java bytecode submissions? They wouldn't even have to change the framework.

[cabbage]

The thing with Java is that it's performance might not be the best. But in programming challenges it shouldn't matter as the better algorithm always outmatches the better performing language.

Just kidding, but I hope Java won't be much of a change from C++ so I can participate as well.


ptsp looks very interesting, thanks for sharing!

[delt0r]

If your coding java for the first time from C++ you probably will have performance issues. Not because java is slow, but because java is not C++. So you can't write like C++ really. Otherwise there is no better way to learn other than by doing. As a rule of thumb you should be about within a factor of 2 of C/C++ code performance wise for "general algorithms" or general computation. In my cases i in fact have better performance than C/C++ code it replaces. But that is a bit of effort. For some raw, but fairly simple number crunching like small matrix multiples or small FFTs, then C/asm is going to be much faster.

Unfortunately I think that C++ to java is quite a change really. Just think that every object is by reference/pointer and that every method is virtual, with no multiple inheritance. Also generics are *not* templates, so don't expect that stuff to work the way you want. Method call overhead is very low (recursion can be faster than loops) and short lived objects are also pretty efficient speed wise with the JIT.

Ok so that was more or less OT. meh is pretty quite here right now anyways.

[drobles]

Hi guys,

I'm one of the guys running the Ms. Pac-Man vs Ghosts competition, mainly working in the website.

What's with academics and Java?

This competition is run just by two: a postdoc and a phd student. Obviously we're not working on this full time (this is just a hobby), and we don't have the time nor the knowledge to open the competition to 20 programming languages between the two of us. In other words, we don't have the power of the aichallenge in servers or number of people. In fact, we just have one server at the moment for running the games! Plus the website hosted in Heroku. If more people wanted to help us we would be more than happy to allow more programming languages for the next iteration of the competition (this is run 2-3 times a year).

We decided to use just one programming language. I don't love Java, I don't hate it neither, but unfortunately is probably the most popular language in most universities and this is an academic competition (IEEE conference), so this is why we chose it. Correct me if I'm wrong, but I think most universities now start teaching Python and Java, and Python is much slower than Java (not that Java is fast, but is not as slow as Python, and in this kind of games tree search speed is important), so this is why we went with Java. However, by looking at the rankings of the aichallenge, I can see that C/C++ might have been a better option! I don't know if that's because C/C++ is more popular in the aichallenge, or because the best programmers do C/C++ and ended up in top of the rankings.

Would it kill them to allow java bytecode submissions?

You're right! JRuby, Jython, Scala, Clojure would be easy to integrate! But again, we don't have too much time right now to add new stuff. But if anyone wants to help you are more than welcome!

Anyway, they definitely need a better visualizer.

I totally agree! I wrote the javascript replays, and to be honest it was my first time writing something like this in Javascript, and it definitely can be improved. The game state representation is ultra heavy, but we'll improve it for the next iteration.

But c'mon guys, the algorithm and the heuristics are more important in this type of competitions, don't be Java haters and submit a controller!

BTW, when is the next aichallenge? Do you guys know the game that will be used?

[Memetix]

Just to let everyone know, the PacMan vs Ghosts competition is about to end, submissions closed about 3 weeks ago and we are in the final week of play-offs. I know some people from the last aichallenge joined in and came up with some great bots. The level this year was a lot higher with the best pacman bots averaging about 90000 points (compared to a winning score of 40000 in the previous competition). Even so, the best ghost teams also managed scores about 20% better than last year.

The good news is that the next competition will start in about a weeks time and if you want to join in, you can download the software and start programming. There have been several good ideas for improvements to the match-up software and tweaks to the game scoring. As yet it has not been announced which of these ideas will be implemented.

[PeterV]

While waiting for the new aichallenge, this is also a nice contest I've signed up for: http://www.codecup.nl
They support Pascal, C, C++, Java, Python, Haskell or Javascript
The final deadline is January 12, 2013, so time enough left to do some coding