随机播种 [英] Random Seeding
问题描述
如果你只为随机数发生器使用32位种子，
这意味着你只能产生最大值
2 ^ 32（约4十亿）不同的序列？
Mersenne Twister怎么样呢？它的周期很长
of 2 ^ 19937-1。如果你只使用32位种子，你是否只能获得这个数字环的一小部分？
这组序列是否仅限于
期间的开始，即你的序列必须从某个地方开始
之间的指数1到40亿（忽略
之后所有可能的数据）?.
或者我误解了种子的概念？
（我正在使用来自...的原始c代码 http://www.math.sci.hiroshima-u.ac.j...at/MT/emt.html ）
If you only use a 32 bit seed for a random number generator,
does that mean you can only ever produce a maximum of
2^32 (approx 4 billion) different sequences?
What about the Mersenne Twister, with it''s massive period
of 2^19937-1. Will you only ever have access to a tiny
portion of this ring of numbers, if you only use a 32-bit seed?
Will this set of sequences be confined to the beginning of the
period, ie, your sequence will have to start at some place
between index number 1 and 4 billion of the period (ignores
all the possible data after this)?.
Or have I misunderstood the concept of a seed?
(I am using the original c code from...
http://www.math.sci.hiroshima-u.ac.j...at/MT/emt.html)
推荐答案
我建议你可能得到2 ^ 32个序列，其中2 ^ 19937-1
数
元素每个。
I would suggest that your possibly get 2^32 sequences with 2^19937-1
number
of elements in each.
> （在此之后忽略所有可能的数据）？
这应该是：
（忽略所有可能的起点）在此之后）？
> (ignores all the possible data after this)?.
this should have read:
(ignores all the possible "starting points" after this)?.
我建议您可能获得2 ^ 32个序列，每个序列中包含2 ^ 19937-1
元素。
I would suggest that your possibly get 2^32 sequences with 2^19937-1
number of elements in each.
这就是我的想法。这看起来相当有限。
然而，我刚刚意识到Mersenne Twister c代码
由它的发明者提供更长的种子。
"那些需要初始种子超过32位
长度的人可以使用init_by_array（）进行初始化，
允许任意长度的数组为种子
我想你可以使用随时间播种的阵列，
进程ID，以及来自
$ b $的压缩信息b / dev / random或/ dev / audio。我仍然认为它会是非常难以保证起点
在数字环上均匀分布？
That''s kind of what I thought. It seems fairly limited.
However, I just realised the Mersenne Twister c code
provided by it''s inventors allows for longer seeds.
"Those who need initial seed with more than 32-bit
length may use init_by_array() for initialization which
admits an array of arbitrary length as seeds"
I guess you could use an array seeded with the time,
the process ID, and some compressed info from
/dev/random or /dev/audio. Still I imagine it would
still be very difficult to guarantee starting points
distibuted uniformly around the ring of numbers?
>如果你只为一个随机数生成器使用32位种子，
>If you only use a 32 bit seed for a random number generator,
这意味着你只能生成最多的
2 ^ 32（约40亿）不同的序列？
是的，假设您总是在播种后生成第一个数字
的序列。可以使用二级
种子。通过使用一个来获取和丢弃N个伪随机数，然后使用一个，这取决于发生器，可能会也可能不会改善
任何东西。
Mersenne Twister怎么样，它的时间很长，是^ ^ 19937-1。如果你只使用32位种子，你是否只能访问这个数字环的一小部分？
很可能瘫痪一个好的伪随机数发生器
种子很差。
这套序列被限制在
期间的开头，即你的序列必须从某个地方开始，在指数1到40亿之间（忽略
所有可能的数据）这个）？。
does that mean you can only ever produce a maximum of
2^32 (approx 4 billion) different sequences?
Yes, assuming you always start a sequence with the first number
generated after seeding. It may be possible to use a "secondary
seed" by getting and throwing away N pseudo-random numbers before
using one, which, depending on the generator, may or may not improve
anything.
What about the Mersenne Twister, with it''s massive period
of 2^19937-1. Will you only ever have access to a tiny
portion of this ring of numbers, if you only use a 32-bit seed?
It is quite possible to cripple a good pseudo-random number generator
with a poor seed.
Will this set of sequences be confined to the beginning of the
period, ie, your sequence will have to start at some place
between index number 1 and 4 billion of the period (ignores
all the possible data after this)?.
我认为没有任何保证，这将取决于发电机上的
。
Gordon L. Burditt
I don''t think there''s any guarantee of that, and it would depend
on the generator.
Gordon L. Burditt
这篇关于随机播种的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！
