Re: into computers or random number theory (please read this)

Posted by Dale on Apr 04, 2002 at 10:51

Re: into computers or random number theory (please read this) (y)

has anybody developed a proccessor which can produce random numbers(NOT PSUEDO RANDOM, OR IN NEED OF A SEED)but truly random.

It is, by definition, impossible to create an algorithm that produces random numbers. Any algorithm that you create can be executed a second time with the same starting point and events to yield the same sequence. A truly random numbers sequence requires a truly random source. Radioactive decay is about the closest feasible solution we have for really random. Microsecond time resolution of user input as a seed is good enough for most processes.

By coincidence, yesterday I tested a system that was designed to produce random motion. Everyone observing the tests agreed that the motion was random yet I knew the algorithm that I used was far from random. It would just take days to repeat the pattern so, to a human (or even a simple computer analysis), it appears totally random. That was all that was needed and it took 3 lines of pseudo-code.


You need to be careful, however, since many of the tests for randomness that are currently used seem totally incorrect, IMHO. For example, is the sequence 7,7,7,7,7,7,7,7,7,7,7 the product of a good random number generator? All of the tests I have ever seen would say no, however, by definition, random means unpredictable. That sequence is highly unlikely from a random process but nothing says that it is predictable. In fact, if you assume that anyone creating a random number generator would make sure it didn’t produce such a sequence, it is actually less predictable than a more “random” sequence. In other words, a truly random number generator’s output is indistinguishable from a predefined infinite length sequence.

Follow Ups:

Post a Followup



[ Forum ] [ New Message ]