wu :: forums - Print Page


    
      
        wu :: forums
        (http://www.ocf.berkeley.edu/~wwu/cgi-bin/yabb/YaBB.cgi)
      

        general >> wanted >> Not-totally random set generation
        
(Message started by: hoogle on Dec 10^th, 2008, 12:05am)

Title: Not-totally random set generation
Post by hoogle on Dec 10^th, 2008, 12:05am

I would like generate a data set that is not totally random. I have an idea . Please comment about its flaws. If you have any other function suggest that too.

Suppose A is an array, of size n, that needs to be filled with numbers. As I told you it shouldn't be totally random.

The function I thought of is follows:

A[i]= rand(0,1) + slope*(i) ; 0<=i<=n and slope<1/n

Explanation of the function: When slope tends to zero the above function becomes just like random number generator(between 0 &1)

rand(0,1) generates a random number between 0 and 1

This is one of the data sets I am planning to consider in my project.

Can you please tell me the advantages and disadvantages of this function.

Can anyone tell me if they have a better not-totally random set generation function.

Title: Re: Not-totally random set generation
Post by towr on Dec 10^th, 2008, 1:10am

What do you want to do with your dataset?

Personally I'd stick a parameter in front of the "rand(0,1)" as well. It makes sense to take Data+Noise, where you can manipulate the amounts of both. Or perhaps use an overall amplitude, plus a factor for the noise: A(D+f*N)
On the other hand, it may make sense to put noise on the magnitude of the data, A*D*(1+f*N)
Or to combine the two A*(D*(1+f₁*N₁)+f₂*N₂)

But without knowing what you are using the dataset for, it's like trying to recommend tourist attractions without knowing what country you're in.

Title: Re: Not-totally random set generation
Post by hoogle on Dec 10^th, 2008, 7:31am

Quote:

What do you want to do with your dataset?

I am working on analysis of parallel sorting algorithms. (Hyperquicksort and parallel merge sort)

I worked on these two input formats (data set):
1) Random data set
2) Fully sorted data set

I also wanted to work on a data set which is not totally random.

@towr
Thanks for your help
Any more comments now

Title: Re: Not-totally random set generation
Post by Grimbal on Dec 10^th, 2008, 7:56am

Another case of partially-random data that can happen in the real world is a set of perfectly sorted elements with a number of random elements appended at the end.

Title: Re: Not-totally random set generation
Post by towr on Dec 10^th, 2008, 11:02am

You could start with a sorted array (or the reverse), then make K random swap, for some value of K (for example, you could swap ~25% of all elements, or just 10 in total)