![]() The neat thing about algorithmic generation is that the contents of this mostrously huge book are not explicitly stored, they are computed as needed (using our position in the book). ![]() In the basics of pseudorandom number generation, we introduced an analogy for algorithmic random number generation-the idea that a random-number generation scheme is like a book of “random numbers”, where the seed for the generator tells us which page to begin reading from. (In contrast RNGs used to construct stream ciphers for secure communication are believed to be infeasible to predict, and are known as cryptographically secure). Surprisingly, the general-purpose random number generators that are in most widespread use are easily predicted. But if you were running a lottery, it might matter a lot. If you're performing a simulation, it may not matter at all. Whether predictability matters depends on the application. ![]() In that sense, it is possible for an entirely predictable random number generator to pass a battery of statistical tests for randomness. Thus, the above numbers both “look random” and are also “totally predictable”. You could predict that if you came back next week and read this page, the exact same numbers will be here, and if someone asked you “What comes after 0x17de4ab5 you could be pretty sure the answer is 0x3fbae22f. But if you come back and read this page tomorrow, they'll be the same and they won't seem quite as random. If you've never seen this page, they ought to look pretty random. To most people, predictability seems like the antithesis of randomness, yet it is in part a matter of perspective. A random number generator is predictable if, after observing some of its “random” output, we can make accurate predictions about what “random values” are coming up next.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |