Generation of uniformly distributed random numbers (0,1) - page 15

New comment
 
joo >>:

И так, какой из трех генераторов - предложенный Вами, определить сможете?


И что значит:"Стандаргный ГПСЧ ЛУЧШЕ чем написанный вами. :)"?

So what's written is MathRand is better. :)

 
SProgrammer >>:

Значит то что написанно - MathRand лучше. :)

That's what I was suggesting with this feature.

 

You need a clear criterion that S voices as 'better' or 'worse'.

By eye, you can see that the points of true randomness are not so clustered.

 

In the first picture the "frequencies" are not evenly distributed, not naturally :) - A lot of high frequency, in the second picture the distribution is not very good, in the third ... Normal white noise. We have to use its definition. Its spectrum is infinite from "-" infinity to "+" infinity, if my memory serves me correctly. :)

 
lea >>:

И на кой нам белый шум, если нужно равномерное распределение?

p.s. Вы в уме делаете двухмерное преобразование Фурье, что утверждаете что-то про спектр? :D

Aaaa.... distribution.... Sorry, I'm off. :)

Yeah, I can do it in my head, :)

Check ... :)))))))

 
lea - Too late, mate :)))
 

These pictures are just one test for uniformity. But you also need the process to have the correct correlation properties. The topic is endless and inexhaustible.

 
Mathemat >>:

...Но нужно и чтобы у процесса были правильные корреляционные свойства. Тема бесконечна и неисчерпаема.

Yes, the topic is interesting, but what do you mean by ' the process has the right correlation properties'?

 

The process defined by the PRNG, of course. What should the process of values (or their differences) be? I don't know yet.

 
Mathemat >>:

У процесса, задаваемого ГПСЧ, конечно. Каким должен быть процесс значений (или их разностей)? Я пока не знаю.

You mean distribution? That's the way it's going to be. Or what do you mean? You can determine the distribution by the frequencies.

1...8910111213141516171819
New comment