Ok if you guys are doing anything but killing time or looking for an interesting slant then read no further... Ok, a couple of questions: My problem is two-fold: first, what, in your opinion is a good random number generator? Of course the number should fall into a range of numbers so it is not truly random but i think you get the point. Next, what algorithm might you employ to (guess|discern) what that number might be? Ok, i know that it is probably ridiculous. I also understand that it is likely impossible. Has anyone ever tried? Would the absence of a valid heuristic keep it from being attemptable? Could you use a GA, and irresponsibly cycle thru generations until chance( and statistical probability) began to score random points? My apologys if this post is a waste of space.(i warned you at the beginning) Also keep in mind that i can forsee no practical application if you could somehow accomplish this. I''m really just wanting to see if someone has a theory on how this might be done or could state simply why this could not be done.( I am of the mind that you would score points randomly without ever evolving a likelyhood greater than statistical probability, but i am open to ideas.) Anyway, feel free to shoot out any nutty idea or comments that you think are pertinent, and thank you for any time you spend considering this. Dreddnafious Maelstrom "If I have seen further, it was by standing on the shoulders of Giants" Sir Isaac Newton

I think you should check out Markov chains as a way of describing repeated sampling from a probability distribution.

As to prediction... if it''s not truly stochastic, then there are methods for modelling the time series of samples taken from the distribution. Check out Bayesian methods for induction.

As to whether it can/has been done... it certainly has. It''s the well trodden path of learning from noisy data.


Cheers,

Timkin

quote: Original post by Dreddnafious Maelstrom
Of course the number should fall into a range of numbers so it is not truly random but i think you get the point.

Why isn't it random if it's in a range? Would you not agree that if you can't predict the next number, even if you know it will be between two given numbers, that it is random?

Also, given an apparatus that generates a random number, there would of course be no way to determine the next number, since it's random.

<opinion type=flammible>
Now, there really is no such thing as randomness, since the universe is deterministic, and everything we perceive as or call random is in reality simply too complex for us to comprehend, though there would be infinite merit in figuring out the overarching functions of the universe.
</opinion>

Later,
ZE.


//email me.//zealouselixir software.//msdn.//n00biez.//
miscellaneous links


[edited by - zealouselixir on January 9, 2003 8:31:36 PM]

So Heisenberg was wrong, eh?

cu,
Prefect

quote: Original post by ZealousElixir

Original post by Dreddnafious Maelstrom
Now, there really is no such thing as randomness, since the universe is deterministic

I think chaos theorists would argue this one.

Wow, great, this topic brought on the exact dialogue that i was hoping for.

@ ZealousElxir, that was in a nutshell my point. perhaps i was mixing chaos theory with the vanilla version of randomness, but for a number to be truly random then i dont imagine it could be contained within a parameterized subset. (a bit abstract but i understand your concrete point)

what got this started was just curiosity. I wondered "could i develop an algorithm that could consistently exceed statistical probability regarding a set of data that should be immune to such prediction."

This lead me to recall a book i had read, (i think it was quoting Sid Meyer) that basically stated that GetTickCount, used as a random seed generator over time would develop predictable patterns, and that more complex methods were necessary to insure a more accurate representation of randomness.

which brings me full circle to my original post. First how to insure that my random number generator does not output a predictable pattern over time. (otherwise the deduction algorithm would simply be identifiying a pattern) Then next, where to begin as far as determining the fitness of any given deduction algorithm. (other than the obvious, "Hey, it got one right!!")
Assume you are dealing in unsigned, integral numbers. if the random number is 8, it doesnt seem like there is really any difference in the deduction 7 or 32512. (or is there?)

Anyway thanks ZE for "getting it" when my post was a bit unclear.
and thanks Timkin for the google keywords, let me educate myself a bit on this and maybe i can come back with something a bit more intelligent to say about it.



Dreddnafious Maelstrom

"If I have seen further, it was by standing on the shoulders of Giants"

Sir Isaac Newton

First, a little conversation...

quote: Original post by Stilton
I think chaos theorists would argue this one.

Yes and no. Chaos theory actually implies that there is no randomness, and that what appears to be random is actually the result of the repeated application of a few simple rules. However, this is nondeterministic even though it isn''t random, since it cannot be predicted without repeatedly applying the rules. Many people, using the "universe as computer" line of thought, hypothesize that there is a limit to the speed of "computation" in the universe, somewhat akin to the maximum velocity, the speed of light, and that therefore we will never be able to construct computers capable of computing those rules faster than the real universe does, and therefore we will never be able to predict - or determine - what will occur until it does.

Is the universe really predictable? Does it follow nice continuous curves? At first the thought is comforting, but then how do you account for free will? Or is there no free will? Chaos theory, or at leaast a little randomness, solves that problem, but implies that we will never be able to completely understand the universe. Is that better?

Personally, I don''t care. Philosophy is probably the least productive way to quickly drive yourself insane.


...and now getting back on topic...
What''s a good random number generator? many people recommend the Mersenne Twister.

How can one find patterns in pseudorandom (ie, chaotic) data? Well, GAs can do just about anything that can be done, so that''s not an absurd idea; GAs just might not get it done within your lifetime. How would you apply GAs? Would you generate expression trees or something?

I think it''s logically impossible for the same reason you can''t compress random data. Think about how many unique symbols (ie. numbers) can be represented by how many bits.

1 bit = 2 symbols
2 bits = 4 symbols
3 bits = 8 symbols
4 bits = 16 symbols
...
n bits = 2^n symbols

It is impossible to use 4 bits for 17 unique symbols. So lets say we have a random bit generator generate 4 bits for us. If it can''t generate all 16 possible outcomes then it isn''t truly random. If it is truly random then we can''t represent the possible output using less than 4 bits. This is true for any number of symbols or length of data. If it happens to generate a highly redundant (ie. compressible) sequence sometimes, the fact remains that it could generate any other possible sequence of that length.

Well, in the finite moments we have, there will always be an infinite matter of choices to an infinite matter of actions to the infinite matter of mistakes...and so on.

I think that actually the best way of puting everything is that:

Time has infinite ways of staying the same.

You see, there will never be true randomness, in the sense that people with above normal intelligence can rationalize. For instance, where does the second count come from, originally? No clue, right? well, when you are in your mother''s womb, there is one sound that is constant, her heart. A pregnant women''s heart beats at a scale or one per second, relatively speaking of course.

Where does this beating occur, and why? Well, in her heart''s muscles'' cells, of course. The pulsating origin of her heart. But, then, why do they beat in such that way?

Well, getting down to very small items, the smallest items of the universe, you will find have a perceptual consistancy to them. Now, I am not talking about using machines, money, resources and human minding to keep track, do years of research and actually put out scientific data so complex and lengthy, that your time reading it would exceed your lifespan, but I am talking about simple observation.

Why would I say all of this? Well, time is based on seconds. Seconds are based on humans. Randomness is based on seconds(in most equations), and if all else is true, humans produce randomness that they themselves say is not true

Okay, forgive the silliness and jokes, but you must realize truly random numbers can never be. Why you ask? Run a test Take a few numbers, make a basic mathematical equation to determine randomness and throw in your data to see what comes out. Now, run this test multiple times(into the thousands)....notice anything? Usually, the numbers that pop out the most are never multiples of three. They are also, usually, never multiples of 5.


Oh, and abstract theories lead way to abstract arguments. Do you realize that we discuss such things to ''see if'' we can do them, while others around us usally have no clue as to what the hell we are talking about. Sometimes, as someone long ago pointed out to me, we come up with conflict and pricey principles to force us to solve those with the same mindset used to create them......a neverending cycle.

Oh and one more thing, if I flip a penny and I cover the penny, is it heads or tails? You know what the outcome can be, but is it random? Two choices, randomness? Yes Well, actually, until you see the results, both choices exist and are actually true, until the choice of the parallel presents itself onto the viewer. In which case, you become the reality and the parallel of that choice Loving Quantum Physics...hehehehhe

Latez peeps,
Hangman, Teamo Supremo

Been a long time since I was reading up on these things...
But in theory if you have an irrational number (say pi) which you can determine at leasure (you have an algorithm) you just simply start somewhere (say the 3342th decimal place), and take the ensuing figures as your random numbers. Sure you have 2 problems. 1st It isn''t truly random as the sequence is predetermined, and 2nd you still need to randomly choose where to begin.


What you get then ist stochiastic behaviour in a controlled system. As long as you view the resulting output in theory things should look random. And it is the behaviour we are interested in n''est-ce pas?