583.txt

This complete matlab for neural network

发信人: GzLi (笑梨), 信区: DataMining
标  题: [合集]matrix calculation
发信站: 南京大学小百合站 (Tue Oct 29 18:48:44 2002), 站内信件

helloboy (hello) 于Thu Oct 24 09:19:06 2002提到：

I am devoting myself into a shortest path solution problem.
In my research, I came up with a hard problem which slow down
my research progress. If you are interested in such a kind of problem
Research together with me. I am anticipating your response.
In my another paper, I use matrix A to represent linkage matrix.
When I calculates M=A+A^2+A^3+.....+A^n+......, I use
(E-A)M=E for fast calculation.
But in my this new article , I redefined the A^2 as
A^2(i,j)=min(A(i,k)+A(k,j))
          k
I also testify that  A^(m+n)=A^m+A^n
So I now need to calculate M=min(A,A^2,A^3,...).
Note that the meaning of A^n is different from the traditional matrix multip
lication.
I can't use E-A any more. What else can I use? I am confused.


fervvac (高远) 于Thu Oct 24 16:23:00 2002提到：

What does your min() means?

According to your definition, it seems A^n contains values x at cell(i,j)
s.t., x is the shortest path from i to j via (n-1) node. Therefore, if
ytour min() is to take min() among all corresponding cells in the matrix
array, then it should be the shortest distance between i, and j and that is
a classic problem.



helloboy (hello) 于Thu Oct 24 19:22:23 2002提到：

min means the corresponding matrix cell's count 


fervvac (高远) 于Thu Oct 24 20:28:05 2002提到：

Then it is the all shortest path problem? If so, just use dijkstra algorithm
and it can be coputed in O(|v|^3) time. (possible improvements exist).



helloboy (hello) 于Fri Oct 25 19:43:38 2002提到：

So I am developing a new way to improve this .