Re: Add terms surrounded by zero together in matrix
- To: mathgroup at smc.vnet.net
- Subject: [mg59180] Re: Add terms surrounded by zero together in matrix
- From: Maxim <ab_def at prontomail.com>
- Date: Sun, 31 Jul 2005 01:30:50 -0400 (EDT)
- References: <dcccur$3j7$1@smc.vnet.net> <dcf34j$lf9$1@smc.vnet.net>
- Sender: owner-wri-mathgroup at wolfram.com
On Sat, 30 Jul 2005 05:27:47 +0000 (UTC), Paul Abbott
<paul at physics.uwa.edu.au> wrote:
> In article <dcccur$3j7$1 at smc.vnet.net>,
> "mchangun at gmail.com" <mchangun at gmail.com> wrote:
>
>> I think this is a rather tough problem to solve. I'm stumped and would
>> really appreciated it if someone can come up with a solution.
>>
>> What i want to do is this. Suppose i have the following matrix:
>>
>> 0 0 0 1 0
>> 0 0 1 2 0
>> 0 0 0 2 1
>> 1 3 0 0 0
>> 0 0 0 0 0
>> 0 0 0 0 0
>> 0 0 1 1 0
>> 5 0 3 0 0
>> 0 0 0 0 0
>> 0 0 0 3 1
>>
>> I'd like to go through it and sum the elements which are surrounded by
>> zeros. So for the above case, an output:
>>
>> [7 4 5 5 4]
>>
>> is required. The order in which the groups surrounded by zero is
>> summed does not matter.
>>
>> The elements are always integers greater than 0.
>
> Have a look at the literature on percolation clusters and in the
> MathGroup archive.
>
> I would have thought that you could use
>
> << Statistics`ClusterAnalysis`
>
> to solve this, but I could not get it to work right. The basic idea I
> had was to locate the non-zero entries in
>
> dat = {{0, 0, 0, 1, 0}, {0, 0, 1, 2, 0}, {0, 0, 0, 2, 1},
> {1, 3, 0, 0, 0}, {0, 0, 0, 0, 0}, {0, 0, 0, 0, 0},
> {0, 0, 1, 1, 0}, {5, 0, 3, 0, 0}, {0, 0, 0, 0, 0},
> {0, 0, 0, 3, 1}};
>
> using Position:
>
> pos = Position[dat, n_Integer /; n > 0]
>
> and then apply FindClusters to these positions. For example,
>
> cl = FindClusters[pos, DistanceFunction -> BrayCurtisDistance]
>
> gives 3 clusters (I tried other distance metrics and several of the
> myriad of options to FindClusters to try to get FindClusters to only
> find "connected" clusters). From these clusters,
>
> Extract[dat, #] & /@ cl
>
> gets the values in each cluster, and
>
> Total /@ %
>
> gives the totals for each cluster. Maybe others can see how to improve
> on this -- or explain why this approach cannot work.
>
> Cheers,
> Paul
>
This can be done as follows:
In[1]:=
<<statistics`
dat = {{0, 0, 0, 1, 0},
{0, 0, 1, 2, 0},
{0, 0, 0, 2, 1},
{1, 3, 0, 0, 0},
{0, 0, 0, 0, 0},
{0, 0, 0, 0, 0},
{0, 0, 1, 1, 0},
{5, 0, 3, 0, 0},
{0, 0, 0, 0, 0},
{0, 0, 0, 3, 1}};
pos = Position[dat, n_Integer /; n > 0];
cl = FindClusters[pos,
DistanceFunction -> (Boole[Norm[# - #2, 1] != 1]&),
Method -> Agglomerate];
Total /@ (Extract[dat, #]&) /@ cl
Out[5]=
{7, 4, 5, 5, 4}
But this algorithm is heuristic in nature and might not give correct
results. Another way is to think of the problem in terms of operations on
graphs:
In[6]:=
<<discretemath`
Module[{L, LL},
L = Position[dat, a_ /; a != 0, {2}, Heads -> False];
LL = Flatten@Position[L, {x_, y_} /; Norm[{x, y} - #, 1] == 1,
{1}, Heads -> False]& /@ L;
LL = ConnectedComponents[FromAdjacencyLists[LL]];
Total@Extract[dat, L[[#]]]& /@ LL
]
Out[7]=
{7, 4, 5, 5, 4}
This assumes that the groups should be 4-connected.
Maxim Rytin
m.r at inbox.ru