Re: Random Math Question

a little while ago i actually had to take a class on creating accurate simulations of systems... which included generating random numbers.

i believe that most computers (and their computer languages) are unable to really come up with a random set of numbers. most of the time most random number sets will require an initial seed to generate the random numbers from. if one is not provided it simply uses the time (epoch?) as the random seed. most random numbers generated from a computer are called pseudo-random because they can either be predicted from the seed, or they will begin to repeat at one point in time.

the only true way to get random numbers would be to have them generated by a physical process that is completely unpredictable (number of raindrops falling or perhaps radioactive decay). Thus it's very hard to actually determine the randomness of an algorithm, you can only hope to make it more random by perhaps randomizing the seed every time, or something to that effect.

some quick searches on comparing pseudo-random numbers to random numbers, or even looking up 'random number generators' should give you a little more insight onto your question.

Comment on Re: Random Math Question

Replies are listed 'Best First'.
Re^2: Random Math Question by Limbic~Region (Chancellor) on Oct 10, 2005 at 21:47 UTC
ickyb0d, Thank you for your response. To be clear, I am assuming we have enough entropy for a "true" randomization of a subset of the list. My question is, with that as a given, how many re-orderings using the process I outlined is required before you have an acceptable fake of the entire list. Additionally, how can you determine the "randomness" of the resulting list post processing? It should be possible to determine how random the resulting list is even if the process to generate it was only pseudo-random. Cheers - L~R	[reply]
Re^3: Random Math Question by ickyb0d (Monk) on Oct 10, 2005 at 22:06 UTC
sorry if i didn't answer your question initially. the thing is... that being random can mean, well... anything. it could mean they could all be the same number, or they could all be completely different numbers for each iteration. my guess is that you are interested in having different numbers. in this case you would want to determine the random variance of the set of numbers. This will tell you how spread out the numbers are. You could also possibly take the correlation of all of your points (plotted on a graph). The lower the correlation the less related each iteration is to one another; but i'm not really sure if that's a good measure of randomness. sorry if i still didn't answer your question, statistics never really was my strong point. i still think your best bet is to seek out other algorithms and see which one works best. if you don't have a lot of math background, it might be tough deciphering all the equations. but anyways, i hope this gives you some more insight about all of this.	[reply]