Oh, I've done some work with back-propigation neural nets. They can't learn in-game, and must be trained before hand. I can easily make one to control a ship, but the problem is training it. I've only done simple things, so I may or may not be able pull it off. Training involves simulating the game from the nets perspective and using randomly generated 'genes' and a small mutation rate teamed with cross breeding of the fitist nets all running in a giant genetic algorithm. Several fitness levels can be obtained for different difficulties. IF I can pull it off. That's a big if.

crikey
shows how much i know about programming. i imagined programming AI to be more like

[SOURCECODE]IF FireballIsInLineOfSight
THEN Shoot
ELSE MoveLeftOrRight[/SOURCECODE]

except obviously it would be thousands of lines of code

That's called a Finite State Machine. That may be what we'll be forced into using. It would be a real pain though =)

Neural Nets really aren't as complicated as they sound. They're really just mathematical datastructures. A really big equation that takes in several variables, does a bunch of stuff, and spits out some more variables (like if it shoud shoot, move, which way to move, etc.). They're also quite fast as compared to other AI systems. Its the complex stuff, like nets that can learn in game that make my eyes grow large

I'd be fine to be part of a three person team, as long as, as Dave says, some sort of boundaries of who's working on what are established early on, so work and our efforts aren't unnecessarily duplicatedÖ though I'll state now that I'm pretty certain Dave should be working on the AIÖ Dave seems slightly over-qualified for this project

I've done plenty of work with OGL, but hardly any with SDLÖ though if this is intended for entry into uDG2k3, I've got a week or so to learn in, which should be plenty of timeÖ

On that point, erazorhead, Dave, any thoughts on whether this is to be entered in uDG2k3 or notÖ? I know erazorhead has already stated a preference for a shareware or donateware release, which probably limits the features you'd want to implement in the uDG2k3 version, as an incentive to register the version released post-uDG2k3Ö?

honestly the "-ware" classification is not something i'm really attached to. for ud2k3, i agree; donationware would work best (so we wouldn't be delivering a crippled version to the contest). my motive isn't really profits, it's seeing this concept developed into a good game for the mac. although if cash comes my way as a result, i won't be turning it down

Well, I'll be happy to handle the AI, but no promises on how cool it will be.

Hmm, what's left?

Network code
Physics/Gameplay code
Display w/OpenGL
Sound
Input
General Interface, File saving etc.


I'm fine working on any of these, although it should be noted that this will be my first *real* project using SDL =).

We can, of course, use something other than SDL, but I don't know how to use ObjC or any non-ASCI C/C++ APIs other than OpenGL and SDL, so those are my limitations at the moment.

Quote:Hmm, what's left?

Network code
Physics/Gameplay code
Display w/OpenGL
Sound
Input
General Interface, File saving etc.

I've no experience of network coding, so I'm afraid I'll have to leave that aspect to youÖ

I'll volunteer to cover the input, graphics, as well as file handling aspects of the coding, if you've no objectionsÖ? Though I realise that this is only the most superficial breakdown of the aspects which would need to be codedÖ

Sounds good to me. I'll handle everything else =).

We'll see how things go, with me going away for a bit less than a month soon...

We probably need some sort of CVS system...

can we communicate over AIM while we work? that's collaborative tool #1 for me - being able to share what you're working on as you work on it, get feedback instantly, it works well. do either/both of you have AIM/iChat?

also, i'm sure everyone here knows except me, since you guys talk in secret code , but are we looking at a classic, carbon, or cocoa game here? i spend essentially 0 time in os 9 anymore, so my preference would be carbon or cocoa, but you guys are the geniuses

I don't have iChat as I've still only got X.1.5, although I've got MSN Messenger, otherwise IRC's always an option I supposeÖ but I'm just dl'ing AIM now anywayÖ

As to the other query, that depends on whether we choose SDL (which would be Cocoa) or pure OGL (which would be Carbon.)

With SDL we gain easier portability, as well as a few facilities for easier net, sound, etc. coding, but I'm not as familiar with the API as DaveÖ also, I'm under the (perhaps false?) impression that SDL is slower than pure Carbon or Cocoa OGL codeÖ

With pure OGL we lose the easier portability, as well as the facilities, but I'm at home with the APIÖ unfortunately Dave states he isn't familar the API's which would provide similar net, sound, etc. facilities as SDLÖ

IRC is good for me too, but I can grab AIM.

By pure OpenGL, do you mean using GLUT? Or AGL + a carbon event loop?

By pure OGL I mean AGL with Carbon for eventsÖ

you guys still interested? like it says in my profile... the AIM screen name is erazorheadncf

otherwise post here... let me know the project still lives!

The project still livesÖ though if this is for uDG2k3, we shouldn't really start development on the project until Carlos opens the contest.