How can I speed up this indexing code?

I have a cell array in which each cell contains a string. E.g....

a={'AAAAAA';'BBBBBB';'CCCCCC';'AAAAAA';'DDDDDD'};

Each cell in array a is associated with a row in a numerical array that contains 10 columns. E.g....

b=[0,0,0,1,1,0,0,0,0,0;
   0,1,0,0,0,0,0,0,0,0;
   1,0,1,0,1,0,2,0,0,0;
   3,0,0,0,0,0,0,0,0,1;
   0,0,0,0,0,0,2,0,0,1];

Some strings in a are repeated such as 'AAAAAA' as shown above. What I need to do is find all repeated cells in a and sum the assocated columns from b into a single row. This should result in two new arrays (unia and bnew) which have equal numbers of rows but every string in unia is unique.

Easy enough to do with a loop such as:

unia=unique(a);
bnew=zeros(numel(unia),10);

for n=1:numel(unia)
    pos=find(strcmp(a,unia{n}));
    bnew(n,:)=sum(b(pos,:),1);
end

This works fine for small arrays but I have a case where a has 6 million cells and unia has 300,000 cells so I need something much faster. Any ideas?

Thanks!

Answer 1

[Ive J]

Avoid comparing strings within the loop and instead take advantage of the index vector from unique:

a = ["A", "B2", "A", "C", "AA", "B2", "B2"]; % use strings instead of cell array of characters, they're much more efficinet to work with
b = randi([0 2], numel(a), 3)

b = 7×3

1 0 1 0 1 2 0 1 2 0 2 1 2 0 2 2 2 0 0 2 1

[anew, ~, idx] = unique(a);
bnew = arrayfun(@(x) sum(b(x == idx, :), 1), 1:numel(anew), 'uni', false);
bnew = vertcat(bnew{:})

bnew = 4×3

1 1 3 2 0 2 2 5 3 0 2 1

anew

anew = 1×4 string array
    "A"    "AA"    "B2"    "C"

Also, you can use tall arrays when dealing with large arrays.

[bhousden]

Perfect! Thanks for your help.