Search All of the Math Forum:
Views expressed in these public forums are not endorsed by
Drexel University or The Math Forum.



Re: associating sequences of numbers
Posted:
Jul 20, 2011 12:58 AM


On Tue, 19 Jul 2011, checker wrote:
> I'd like some guidance on how to associate series of numbers. > If I have several series of whole numbers > > A = 2 4 6 1 8 4 4 5 1 > B = 1 1 8 4 4 7 3 > C = 8 3 2 2 5 6 4 4 2 5 > > That are represented in an array as > > Set1 = > 2 4 6 1 8 4 4 5 1  > 1 1 8 4 4 7 3    > 8 3 2 2 5 6 4 4 2 5 > > And I'm presented a second array like > > Set2 = > 2 4 6 1 8 4 4 5 1  > 1 1 8 4 5 7 3    > 8 3 2 4 4 2 5    > 5 8 8 9 3 2 3 3 1 3 > > where > Set2(2,5) changed to 5 > Set2(3,4:6) removed > Set2(4,:) added
Huh? Does that mean anything?
> I'd like an approach that would associate > Set1 row1 with Set2 row 1 with high probability > Set1 row2 with Set2 row 2 with high probability > Set1 row3 with Set2 row 3 with high probability > and all other associations (say Set1 row 1 with Set2 row 2) with low > probability. > > The important information is contained in a row and is represented by > the value and the sequence. That is, 1 2 3 would be poorly associated > with 3 2 1. The row location in the array is not important. That is, > the first row in the first array could be duplicated in the 2nd row of > the second array. In this case, the method would associated the first > row of the first array with the second row of the second array with high > likelihood.
Confusing in view of the previous paragraph.
> It looks like a stats problem, but the importance of the sequence (the > relation of one number to it's neighbor) is undermining my meager > understanding.
If a1, a2,.. a_n and b1, b2,.. b_n are two data sets, one measure of similitude is
(sum(j=1,n) (aj  bj)^2) / n
> Thanks for your comments. > > Chris >



