Deleting rows of repeated labels in Matlab

Question

In Matlab, I have a cell of which the first column is composed of markers of which there are too many. We want to delete the entire row of the cell of these extraneous markers.

We have four marker types within the first column of our cell - '71', '72', '73' and '74': stored as strings. Whenever there is an occurrence of any of these markers, they are repeated ('71' '72' 23 times and '73' '74' 16 times). We want to keep the first presentation each time that markers occurs. Then for '71' '72' keep every 6th, whereas for '73' '74' keep every 4th.

The cell has length 550 so the blocks of 71, 72, 73 and 74 are repeated many times in the cell and there are other markers besides these four present as well that we want to keep.

A small part of the cell is shown below:

'12'    14737   29472
'31'    44747   89492
'4'     44771   89540
'53'    53756   107510
'73'    53831   107660
'73'    54082   108162
'73'    54331   108660
'73'    54582   109162
'73'    56081   112160
'73'    56331   112660
'73'    56581   113160
'73'    56831   113660
'73'    58330   116658
'73'    58580   117158
'73'    58829   117656
'73'    59079   118156
'73'    60579   121156
'73'    60829   121656
'73'    61079   122156
'73'    61329   122656
'63'    351340  702678
'4' 351361  702720
'54'    360342  720682
'74'    360375  720748
'74'    360633  721264
'74'    360883  721764
'74'    361133  722264
'74'    362632  725262
'74'    362882  725762
'74'    363132  726262
'74'    363382  726762
'74'    364881  729760
'74'    365131  730260
'74'    365381  730760
'74'    365631  731260
'74'    367130  734258
'74'    367380  734758
'74'    367630  735258
'74'    367880  735758
'64'    369374  738746
'4' 369379  738756
'51'    378376  756750
'71'    378409  756816
'71'    378584  757166
'71'    378750  757498
'71'    378917  757832
 **continued**

I'm not sure how to do this, I tried to use indices to approach this problem but have so far been unsuccessful.

Could anyone help?

Thanks

What I've tried so far:

function Y = correctRows(y)
Y.type = {};
Y.latency = [];
Y.latency_ms = [];


for i = 1:length(y)

if strcmp(y(i).type,'71') || strcmp(y(i).type,'72')
   Y(i) = y(i); 
    i = i+5;

elseif strcmp(y(i).type,'73') || strcmp(y(i).type,'74')
    Y(i) = y(i); 
    i = i+3;
else
    Y(ii) = y(i);
end 
end

Can you share your code which you developed so far? And an output example — Irreducible, Oct 12 '17 at 13:59
`71` and `72` will never follow each other ? same question for *73* and `74` — Irreducible, Oct 12 '17 at 14:08
@Irreducible they are all randomised, any of them can follow the other — Maheen Siddiqui, Oct 12 '17 at 14:08
You wrote "*We want to keep the first presentation each time that markers occurs."* and then in the next line you self-contradicted your statement with these words: *"Then for '71' '72' keep every 6th, whereas for '73' '74' keep every 4th."* — Sardar Usama, Oct 12 '17 at 15:22
Is there a system or a map of some sort, for the "how-manieth" you want to keep for each label? — Rody Oldenhuis, Oct 12 '17 at 15:30
Need to keep the first and then either every 6th or every 4th depending on the marker — Maheen Siddiqui, Oct 12 '17 at 15:32
Yes, it needs to have the same order as in the original file - the other columns are times and we need the right timing — Maheen Siddiqui, Oct 12 '17 at 15:43
What about other markers, e.g; 4 also occurs multiple times, what would be its criterion? — Sardar Usama, Oct 12 '17 at 16:12

Irreducible · Answer 1 · 2017-10-12T15:25:56.390

0

Here an example with a numerical array instead of a string array.

I think you get the transfer to strings yourself. The main idea of yours was not bad. But you did not define ii and I would rather use a while instead of a for loop.

This code only demonstrates how to cancel certain rows in a numerical Vector. The transfer to any other matrix / cell should be handled individually

And now the code

counter=1;
Result=zeros(size(A));
i=1;
while k<=length(A)

if (A(k)==71 || A(k)==72)
    Result(counter,:) = A(i);
    k = k+6;
    counter=counter+1;
elseif (A(k)==73 || A(k)==74)
    Result(counter,:) = A(k);

    k = k+4;
    counter=counter+1;
else
    Result(counter,:) = A(k);
    k = k+1;
    counter=counter+1;
end

end
%delet unused rows
Results(counter:end)=[];

edited Oct 12 '17 at 15:25

answered Oct 12 '17 at 14:27

Irreducible

864
11
24

Thanks, this works with an array but not a cell. It gives an error for me as I have a cell not an array... – Maheen Siddiqui Oct 12 '17 at 14:38
Is the output also a cell? and which error are you getting? – Irreducible Oct 12 '17 at 14:40
This is the error: Undefined operator '==' for input arguments of type 'cell'. – Maheen Siddiqui Oct 12 '17 at 14:43
2

In your question please add one example of a cell output of your data – Irreducible Oct 12 '17 at 14:49
I don't understand what you mean... I have provided an example of my cell, that's what I need to be able to use. – Maheen Siddiqui Oct 12 '17 at 14:58
Maheen, you should know yourself that the first entry in your cell is a string, so your comparison should be something like this `...strcmp(c{1,1},'71')...` assuming that your data is stored in the variable `c` – Irreducible Oct 12 '17 at 15:19
Increasing array size continuously is not a good idea in terms of performance and memory use. Also `i` as a variable name--> not a good idea (reserved for imaginary numbers) – Sardar Usama Oct 12 '17 at 15:25

score 0 · Answer 2 · answered Oct 12 '17 at 16:20

I'd go for a scalable approach using a generic map (A is your cell):

%       markers      how-manieth   block size
map = { {'71' '72'}  6             23
        {'73' '74'}  4             16 };

for ii = 1:size(map,1)

    % Extract block
    blk  = cellfun(@(x) any(strcmp(x,map{ii,1})), A(:,1));

    % Find the how-manieth entry in each block
    inds = (mod(cumsum(blk), map{ii,3}) == map{ii,2});

    % Empty the rest
    A(xor(blk, inds),:)  = {[]};

end

% Remove all entries marked for removal
A = A(~cellfun('isempty',A(:,1)),:);

If there are also other keys that may be repeated, of which you'd need to keep only the first, then this may be more readable:

%       markers      how-manieth 
map = { {'71' '72'}, 6
        {'73' '74'}, 4 };

% The routine
output = cell(size(A));
index  = 1;

repetitions = 0;
repcount    = 0;
have_match  = false;

previous = '';

for ii = 1:size(A,1)

    % Current label
    current = A(ii,1);

    % If this label equals the previous one
    if strcmp(current, previous)

        repetitions = repetitions + 1;

        % First entry: 
        if repetitions==1

            % Check for mapped labels
            match = cellfun(@(x)any(strcmp(current,x)), map(:,1));
            if any(match)      
                % Mapped label found: 
                index      = index-1;
                repcount   = map{match,2};
                have_match = true;
            end

        % Subsequent entries
        elseif have_match && repetitions==repcount-1        
            output(index,:) = A(ii,:);
            index           = index + 1;
        end

    % Current label is not repeated; just insert the new data
    else
        have_match   = false;
        repcount     = 0;
        repetitions  = 0;

        output(index,:) = A(ii,:); 
        index           = index + 1;
    end

    previous = current;    
end

% Chop-off redundant entries
output(index:end,:) = [];

Deleting rows of repeated labels in Matlab

2 Answers2