create matrices based on a label sequence

Hello,

Please help me with the following:

Consider a 100x10 matrix, called A and a 100x1 vector, called B, that contains labels that refer to the rows of A.

These labels seperate the rows of A, actually they are labels generated by a clustering process.

For example if the number of labels is 3 (and is randomly distributed in B), then all the rows of A are separated in 3 clusters.

With the following commands, we have 3 separate matrices, 1 matrix per cluster:

Cluster1=A(B==1,:);

Cluster2=A(B==2,:);

Cluster3=A(B==3,:);

How can I form different matrices that refer to the different labels avoiding writing 1 command for 1 cluster (maybe with a "for" loop)?

For example, if we have 12 clusters (labels of B), we should avoid writing

cluster1=...

cluster2=...

. .

.

cluster12=...

and the 12 separate matrices (cluster1,...,cluster12) automatically generated.

Thank you very much.

Best,

Pavlos

Answer 1

[Andrei Bobrov]

claster_n = accumarray(B,1:numel(B),[],@(x){A(x,:)});

or with arrayfun

claster_n = arrayfun(@(x)A(x == B,:),(1:max(B))','un',0);

and with for..end loop

n = max(B);
claster_n = cell(n,1);
for jj = 1:n
    claster_n{jj} = A(jj == B,:);
end