Thanks for the explanation Don! It looks like I need better math-foo to
completely understand your explanation...
On Tuesday, February 11, 2014 10:07:05 AM UTC-8, Don wrote:
>
>
> It can be shown mathematically that if you use a multiplier A such that
> A*2^15-1 and A*2^16-1 are both prime, you
It can be shown mathematically that if you use a multiplier A such that
A*2^15-1 and A*2^16-1 are both prime, you get a sequence with period
A*2^15. With A=63663 you get a sequence of nearly two billion. If A was
larger than 2^16, the multiplication would result in overflow in some
cases, so t
:) true... that was silly of me.. I was too quick to post... anyway, the
point I was trying to understand was why this specific combination of
operations would produce good randomness... Running the original solution
you posted a million times (with the max limit set to 1000), below is the
freq
https://groups.google.com/forum/#!original/comp.lang.java.programmer/Z8lfFIDhS7k/wbPLDldPSSgJ
The Multiply With Carry generator is ok for some purposes. There are
generators with longer periods and better statistical properties, but they
are more complicated. The link I provided has code for a 6
@don : awesome +1 . I was not aware of George Marsaglia's Multiply
With Carry generator. But it is soo clll . If you have any good link
for its proof or explanation of the idea behind it then please mention it.
I never knew generating random number can be few lines of code :)
Thanks :)
A random generator will not always (or even usually) produce a permutation
of the possible values before repeating.
If you call my function 10 million times and tally up the results, you will
get a distribution with about a million of each value, as close as you
would expect for a random sample
Because if you put that in a loop you will get a series like:
3 3 3 3 3 3 3 3 4 4 4 4 4 4 4 4 4 4 4 4 5 5 5 5 5 5 5 5 5 5...
On Thursday, February 6, 2014 8:27:27 PM UTC-5, SVIX wrote:
>
> Just curious... Why is this more random than
> (getSystemTimeInMilliseconds() % 10)
>
> On Thursday, Februa
@don: algo look fine..i tested it ,but it did not generate 1-10 with
probabilty of 1/10 everytime.
actually question was asked in an intrview.
question started with displaying all elements in an array randomly without
repetition i.e only once( probabilty 1/10)...which can be done with fisher
yates
Just curious... Why is this more random than (getSystemTimeInMilliseconds()
% 10)
On Thursday, February 6, 2014 1:47:15 PM UTC-8, Don wrote:
>
> Just noticed that you asked for 1-10, and my code will clearly generate
> 0-9. Add one for the desired range.
> Don
>
> On Wednesday, February 5, 2014
Just noticed that you asked for 1-10, and my code will clearly generate
0-9. Add one for the desired range.
Don
On Wednesday, February 5, 2014 4:29:26 PM UTC-5, Don wrote:
>
> // Using George Marsaglia's Multiply With Carry generator
> int rnd10()
> {
>static unsigned int x = time(0);
>x
// Using George Marsaglia's Multiply With Carry generator
int rnd10()
{
static unsigned int x = time(0);
x = 63663 * (x&65535) + (x>>16);
return (x & 65535) % 10;
}
On Sunday, February 2, 2014 2:51:47 AM UTC-5, atul007 wrote:
>
> Generate random number form 1 - 10 with probability of 1/10
@Karthikeya: You accept that your for-loop generates a random integer
between 0 and 2^d - 1, right? Then take a look at
http://en.wikipedia.org/wiki/Rejection_sampling. Basically, you reject
any random integer that exceeds b-a. The probability that you reject a
number is 1 - (b - a + 1) / (2^d - 1)
12 matches
Mail list logo
