Removing adjacent duplicate numbers in rows of a matrix

Hi,

assume the following matrix format:

xx=[1 2 3 3 4 4; 1 1 2 2 3 3 ; 5 5 5 3 3 2]

I want to remove the adjacent duplicate numbers in each row of this matrix so I get the following output:

result=[1 2 3 4; 1 2 3 ; 5 3 2]

Help would be very much appreciated with a formula. Thanks.

Answer 1

[infinity]

Hello,

In your problem, you may get error since your result is not a matrix (number of colume of the first row is different with other rows).

So, you result should be a cell like this

clear

xx=[1 2 3 3 4 4; 1 1 2 2 3 3 ; 5 5 5 3 3 2]

for i = 1:size(xx,1)
 yy{i} = unique(xx(i,:));
end

[AA]

Hi,

but I want the result to be a matrix.

[Rik]

That is not possible in Matlab. Arrays must be rectangular, so every row must have the same number of columns. You must either pad each row with some value (e.g. NaN), or use a cell array.

[infinity]

Hello @AA

Like @Rik said it is not allowed in Matlab. However, we can pass through it with small trick.

Let say, we can create the matrix that has the same number of rows and put the value of elements that are not in "xx" by a fake number, for example 0 or 1000.

Here is a reference for you

clear

xx=[1 2 3 3 4 4; 1 1 2 2 3 3 ; 5 5 5 3 3 2]

yy = 0*xx;
for i = 1:size(xx,1)
    temp = unique(xx(i,:));
 yy(i,1:length(temp)) = temp;
end

[AA]

thank you but the problem is removing the zeroes after that. how can i do that?

what about the diff function? calculating the difference between adjacent numbers and then removing the zeroes?

[infinity]

Hello,

What do you mean by removing the zeros?

Is it removing zeros in matrix "yy" to create another matrix?

[AA]

yes removing the zeroes in the yy matrix.

[infinity]

When we remove 0 in "yy", the number of rows of the new matrix should be the same. The smallest size matrix that we can be obtained by

zz =

     1     2     3     4
     1     2     3     0
     2     3     5     0

[AA]

yes that is great but the order of the numbers has changed. How can I still maintain the orders of the numbers? I dont mind the zeroes.

[infinity]

Here you can refer this

clear

xx=[1 2 3 3 4 4; 1 1 2 2 3 3 ; 5 5 5 3 3 2]

yy = 0*xx;
for i = 1:size(xx,1)
    temp = unique(xx(i,:),'stable');
 yy(i,1:length(temp)) = temp;
end

[Rik]

I would suggest using the code below instead. With a NaN you avoid confusion if the last value happens to be a 0.

clear
xx=[1 2 3 3 4 4; 1 1 2 2 3 3 ; 5 5 5 3 3 2]
%yy = zeros(size(xx));
yy=NaN(size(xx))
for row = 1:size(xx,1)
    temp = unique(xx(row,:),'stable');
    yy(row,1:length(temp)) = temp;
end

[AA]

Hi,

somehow I see an error when I modify the xx and increase the set of numbers. Adjacent numbers can repeat like 3 or 2 or 5 but they are not displayed in the output.

clear
xx=[1 2 3 3 4 4 3; 1 1 2 2 3 3 2; 5 5 5 3 3 2 5]
%yy = zeros(size(xx));
yy=NaN(size(xx))
for row = 1:size(xx,1)
    temp = unique(xx(row,:),'stable');
    yy(row,1:length(temp)) = temp;
end

The last digit 3 or 2 or 5 is not shown in the output. Instead i get

1	2	3	4	NaN	NaN	NaN
1	2	3	NaN	NaN	NaN	NaN
5	3	2	NaN	NaN	NaN	NaN

but I should get

1	2	3	4	3	NaN	NaN
1	2	3	2     NaN	NaN	NaN
5	3	2	5      NaN	NaN	NaN

How can I modify the formula?

[infinity]

Hello,

You can modify the code like this

clear

% xx=[1 2 3 3 4 4; 1 1 2 2 3 3 ; 5 5 5 3 3 2]
% xx=[1 2 3 3 4 4; 1 1 2 2 3 3 ; 5 4 5 3 3 2]
xx=[1 2 3 3 4 4 3; 1 1 2 2 3 3 2; 5 5 5 3 3 2 5]
% xx=[1 2 3 3 4 4 3 3; 1 1 2 2 3 3 2 3; 5 5 5 3 3 2 5 5]
n = size(xx,2);
% yy = 0*xx;
yy=NaN(size(xx));
for i = 1:size(xx,1)
    [temp,ia,ib] = unique(xx(i,:),'stable');
    f = diff(ib)~=0;
    idx = [find(f);length(ib)];
    %  yy(i,1:length(temp)) = temp;
    yy(i,1:length(idx)) = xx(i,idx);
end
yy

Answer 2

[Rik]

You can use diff to find any repeats and remove them. The compare to eps is to avoid strange things happening once you put in decimal numbers, which can lead to rounding errors, which is why exp(log(3))==3 returns false.

xx=[1 2 3 3 4 4 3; 1 1 2 2 3 3 2; 5 5 5 3 3 2 5];
yy=NaN(size(xx));
for row = 1:size(xx,1)
    temp = xx(row,:);
    %keep values that are different from the one before
    %L=[true diff(temp)~=0];%line below is safe for float rounding
    L=[true abs(diff(temp))>=(2*eps)];
    yy(row,1:sum(L)) = temp(L);
end
%remove trailing NaN cols:
while all(isnan(yy(:,end)))
    yy(:,end)=[];
end
disp(yy)