Re: Moving from VC++ 6.0 to VC++.NET
Quote:
Originally Posted by
wavering
Well, I didn't come here to argue about the merits or otherwise of Mopeks but after that comment I have little choice but to reply. What I am trying to do here is move a FULLY OPERATIONAL program from VB6 and VC++ 6.0 to .net
You did mention performance in your original post and I consider it a more important problem than the rather trivial porting issue. And obviously you referred to Mopeks in order to draw attention to this project. There's nothing offensive about that.
http://www.mopeks.org/
You state human higher intelligence is about playing out different scenarios in the mind and then pick the best. Certainly, but there is nothing novel about that and when you decide to simulate it with a genetic algorithm it becomes just an ordinary evolutionary approach (despite your attempts to dress it up in a metaphysical garment).
There's lots of research going on in this area. If you missed my link please have a look,
http://www.cgchannel.com/2014/01/coo...on-for-bipeds/
Watch how the creatures first evolve to walk and then learn to run and to avoid objects thrown at them.
I didn't say the 20 years you've spent on this have been wasted. You've obviously learned a lot and have a working program. What I'm saying is that your image of yourself as the leader of a paradigm shift is not grounded in reality. Your project shows a resemblance with alchemy and the search for the philosopher's stone. I've seen it before. People hope that by molding a program long enough it will magically transmute into an elixir of life, in your case show human intelligence. I strongly recommend that instead of just tugging on for another 20 years you consolidate this project in real science and devise a more realistic strategy for the future (see my previous post).
Good luck.
Re: Moving from VC++ 6.0 to VC++.NET
Razzle
The project you highlight looks to me like a perfectly straightforward Genetic Programming project - a computer program evolves over time. That is old hat and the principle has been around for decades. I was writing such programs in the previous century. The first were written in about 1948 eg "Core Wars", in which evolving programs fight each other for supremacy in RAM, or core as it was called in those days
What mopeks does which I believe is original is to use GP to produce a Library of general purpose programs which use each other. So, if an object in mopeks wants to know the distance to another object is evolves a suitable program which will doubtless use the Square Root program it found earlier. And so on.
I believe this is how human intelligence operates - to solve a problem we use a whole raft of procedures which we have previously learned or generated to create a new procedure. So, you start with virtually nothing as a baby and build up a vast library which is stored in memory.
In mopeks, this evolved library can then be used by any arbitrary object in any environment. So the objects could be:
Razzle, wavering, Observer1, Observer2, Observer3 etc
The environment could be:
CodeGuruBoard
The objective could be to send the audience to sleep as soon as possible
Each object can have totally arbitrary properties eg
IQ, experience, grumpiness, persistence and so on
All of the objects interact in the simulator and an optimum approach is produced. There is no computer programming involved - the evolved programs run within mopeks itself. This is clunky and time consuming right now but so were the early planes - no passengers and a range of a few hundred yards.
Here is Flash Gordon homing in on his objective using the gravity slingshot effect:
http://www.mopeks.org/images/form_di...l_animated.gif
Yes, pretty trivial but bear in mind that this uses a whole library of routines generated by mopeks and is totally general purpose. So, in principle, MOPEKS can be used to solve ANY problem which involves the interaction of totally arbitrary objects, with totally arbitrary properties in a totally arbitrary environment.
I am not aware of anybody releasing a fully operational program to do that. And its free ...
Re: Moving from VC++ 6.0 to VC++.NET
Quote:
Originally Posted by
wavering
So, in principle, MOPEKS can be used to solve ANY problem which involves the interaction of totally arbitrary objects, with totally arbitrary properties in a totally arbitrary environment.
You may be inspired by human intelligence but your approach to problem solving is based on the principle of emergence with evolution as the direction force. That's completely general as you say. It's also why GA is a suitable computational strategy for your project.
The big challenge is to devise a parametrization of the participating objects which allows GA to be efficiently applied (compare the moving biped project I linked to). And to develop a theory for this, that is applying science as opposed to just tinkering with a program.
Mopeks should be pretty small and simple. If it's big and complex (as you indicated) you probably are hardwiring problem solving capabilities into Mopeks and that's cheating because it's supposed to emerge by itself. So why do you need to refine Mopeks further?
If your system is able to come up with the variational principle of physics and apply it to solve the differential game problem you referred to, it's quite an achievement. So why don't you drop the metaphysics and write up an ordinary scientific paper?
