Extending ismember to cells - matlab

I have searched the forum and have not found enough info to help me solve this problem.
Consider the set (cell of vectors)
A = {[1],[1 2],[2],[1 2 3],[1 2 3 4],[1 3]}
I want to construct a matrix B that looks like
B = [1 1 0 1 1 1
0 1 0 1 1 0
0 1 1 1 1 0
0 0 0 1 1 0
0 0 0 0 1 0
0 0 0 1 1 1]
The matrix B specifies membership of vectors with respect to each other. That is, the first row looks at the first element in A, [1], and checks if it is a member of the other vectors, placing a 1 if it is a member and a 0 otherwise.
I can do this using two for loops: one over the elements of A, and another nested, for each element of A, that checks membership with respect to every other member of A.
I want to avoid using for loops. Is there a vectorized solution for obtaining B from A?

With cell arrays it's hard to avoid loops, or their cousin cellfun. This is how I would do it:
[ii, jj] = ndgrid(1:numel(A)); % indices of all possible pairs
result = cellfun(#(x,y) all(ismember(x,y)), A(ii), A(jj)); % see if all elements in the
% first are present in the second

Well you asked for it, so here's an almost* vectorized solution using bsxfun and permute -
lens = cellfun('length',A)
vals = [A{:}]
mask = bsxfun(#ge,lens,[1:max(vals)]')
a = nan(size(mask))
a(mask) = vals
matches = bsxfun(#eq,a,permute(a,[3,4,1,2]));
out = bsxfun(#eq,squeeze(sum(any(matches,3),1)),lens(:))
*: Almost because of the use of cellfun at the start with cellfun('length',A), but since its just getting the length of the cells there, so computationally would be negligible .
Also, please note that this approach would use a lot of memory resources, so might not be beneficial, but just respecting the requirements of a vectorized solution as much as possible!

Related

Output a matrix size n x m, 1 when the sum of the indices is even, 0 otherwise

I'm attempting the following as a hobby, not as homework. In Computer Programming with MATLAB: J. Michael Fitpatrick and Akos Ledeczi, there is a practice problem that asks this:
Write a function called alternate that takes two positive integers, n and m, as input arguments (the function does not have to check the format of the input) and returns one matrix as an output argument. Each element of the n-by-m output matrix for which the sum of its indices is even is 1.
All other elements are zero.
A previous problem was similar, and I wrote a very simple function that does what it asks:
function A = alternate(n,m)
A(1:n,1:m)=0;
A(2:2:n,2:2:m)=1;
A(1:2:n,1:2:m)=1;
end
Now my question is, is that good enough? It outputs exactly what it asks for, but it's not checking for the sum. So far we haven't discussed nested if statements or anything of that sort, we just started going over very basic functions. I feel like giving it more functionality would allow it to be recycled better for future use.
Great to see you're learning, step 1 in learning any programming language should be to ensure you always add relevant comments! This helps you, and anyone reading your code. So the first improvement would be this:
function A = alternate(n,m)
% Function to return an n*m matrix, which is 1 when the sum of the indices is even
A(1:n,1:m)=0; % Create the n*m array of zeros
A(2:2:n,2:2:m)=1; % All elements with even row and col indices: even+even=even
A(1:2:n,1:2:m)=1; % All elements with odd row and col indicies: odd+odd=even
end
You can, however, make this more concise (discounting comments), and perhaps more clearly relate to the brief:
function A = alternate(n,m)
% Function to return an n*m matrix, which is 1 when the sum of the indices is even
% Sum of row and col indices. Uses implicit expansion (R2016b+) to form
% a matrix from a row and column array
idx = (1:n).' + (1:m);
% We want 1 when x is even, 0 when odd. mod(x,2) is the opposite, so 1-mod(x,2) works:
A = 1 - mod( idx, 2 );
end
Both functions do the same thing, and it's personal preference (and performance related for large problems) which you should use.
I'd argue that, even without comments, the alternative I've written more clearly does what it says on the tin. You don't have to know the brief to understand you're looking for the even index sums, since I've done the sum and tested if even. Your code requires interpretation.
It can also be written as a one-liner, whereas the indexing approach can't be (as you've done it).
A = 1 - mod( (1:n).' + (1:m), 2 ); % 1 when row + column index is even
Your function works fine and output the desired result, let me propose you an alternative:
function A = alternate(n,m)
A = zeros( n , m ) ; % pre-allocate result (all elements at 0)
[x,y] = meshgrid(1:m,1:n) ; % define a grid of indices
A(mod(x+y,2)==0) = 1 ; % modify elements of "A" whose indices verify the condition
end
Which returns:
>> alternate(4,5)
ans =
1 0 1 0 1
0 1 0 1 0
1 0 1 0 1
0 1 0 1 0
initialisation:
The first line is the equivalent to your first line, but it is the cannonical MATLAB way of creating a new matrix.
It uses the function zeros(n,m).
Note that MATLAB has similar functions to create and preallocate matrices for different types, for examples:
ones(n,m) Create
a matrix of double, size [n,m] with all elements set to 1
nan(n,m) Create a
matrix of double, size [n,m] with all elements set to NaN
false(n,m) Create a
matrix of boolean size [n,m] with all elements set to false
There are several other matrix construction predefined function, some more specialised (like eye), so before trying hard to generate your initial matrix, you can look in the documentation if a specialised function exist for your case.
indices
The second line generate 2 matrices x and y which will be the indices of A. It uses the function meshgrid. For example in the case shown above, x and y look like:
| x = | y = |
| 1 2 3 4 5 | 1 1 1 1 1 |
| 1 2 3 4 5 | 2 2 2 2 2 |
| 1 2 3 4 5 | 3 3 3 3 3 |
| 1 2 3 4 5 | 4 4 4 4 4 |
odd/even indices
To calculate the sum of the indices, it is now trivial in MATLAB, as easy as:
>> x+y
ans =
2 3 4 5 6
3 4 5 6 7
4 5 6 7 8
5 6 7 8 9
Now we just need to know which ones are even. For this we'll use the modulo operator (mod) on this summed matrix:
>> mod(x+y,2)==0
ans =
1 0 1 0 1
0 1 0 1 0
1 0 1 0 1
0 1 0 1 0
This result logical matrix is the same size as A and contain 1 where the sum of the indices is even, and 0 otherwise. We can use this logical matrix to modify only the elements of A which satisfied the condition:
>> A(mod(x+y,2)==0) = 1
A =
1 0 1 0 1
0 1 0 1 0
1 0 1 0 1
0 1 0 1 0
Note that in this case the logical matrix found in the previous step would have been ok since the value to assign to the special indices is 1, which is the same as the numeric representation of true for MATLAB. In case you wanted to assign a different value, but the same indices condition, simply replace the last assignment:
A(mod(x+y,2)==0) = your_target_value ;
I don't like spoiling the learning. So let me just give you some hints.
Matlab is very efficient if you do operations on vectors, not on individual elements. So, why not creating two matrices (e.g. N, M) that holds all the indices? Have a look at the meshgrid() function.
Than you might be able find all positions with an even sum of indices in one line.
Second hint is that the outputs of a logic operation, e.g. B = A==4, yields a logic matrix. You can convert this to a matrix of zeros by using B = double(B).
Have fun!

Comparing Vectors of Different Length

I am trying to compare two vectors of different size. For instance when I run the code below:
A = [1 4 3 7 9];
B = [1 2 3 4 5 6 7 8 9];
myPadded = [A zeros(1,4)];
C = ismember(myPadded,B)
I get the following output:
C = 1 1 1 1 1 0 0 0 0
However, I want an output that will reflect the positions of the compared values, hence, I would like an output that is displayed as follows:
C = 1 0 1 1 0 0 1 0 1
Please, I need some help :)
There are 2 points. First, you are writing the inputs of ismember in the wrong order. Additionally, you do not need to grow your matrix. Simply try ismember(B, A) and you will get what you expect.
The function ismember(myPadded, B) returns a vector the same size of myPadded, indicating if the i-th element of myPadded is present in B.
To get what you want, just invert parameter order: ismember(B, myPadded).
A quick way of doing this is to use logical indexing. This will only work if the last digit of B is included in A.
A = [1 4 3 7 9];
c(A) = 1; % or true.
An assumption here is that you want to subindex a vector 1:N, so that B always is B = 1:N. In case the last digit is not one this is easy to fix. Just remember to return all to its previous state after you are done. It will be 2 rows extra though.
This solution is meant as a special case working on a very common problem.

How to make a general case of inserting ones in any type of matrix, in the non-principal diagonal

The title might be confusing, here's a particular example to explain myself. Also, I'm not sure how do you call the diagonal that starts in (1,2) and goes onward: (2,3) ; (3,4) and so on. Non-principal, non-main diagonal, not sure at all.
3x3 case
-1 1 0
-1 0 1
0 -1 1
4x4 case
-1 1 0 0
-1 0 1 0
-1 0 0 1
0 -1 1 0
0 -1 0 1
0 0 -1 1
So if the original matrix was a 4x4 (or any other size), I am able to make a matrix the size of the second example. I now have to insert the -1 and 1's in this fashion. This means n-1 number of -1's inserted if j=1, and then, a n-1 number of ones in the non-principal diagonal. When this is done, it's the same but for j=2 and the next non-principal diagonal, and so on.
Thing is, I'm thinking all the time about loops, and too many cases arise, because what I want is to be able to do this for any possible dimension, not for a particular case.
But then I saw this post Obtaining opposite diagonal of a matrix in Matlab
With this answer: A(s:s-1:end-1)
And it seems like a much cleaner way of doing it, since my own way (not finished since I'm not able to figure all the cases) has too many conditions. With a sentence like that, I could choose the diagonal, insert ones, and do it as many times as required, depending of the n dimension.
This leaves the problem of inserting the -1's, but I guess I could manage something.
It seems to mee that you want to obtain the following matrix B of size n × (n-1)*n/2
n = 4;
idx = fliplr(fullfact([n n]));
idx(diff(idx')<=0,:) = [];
m = size(idx,1);
B = zeros(m,n);
B(sub2ind(size(B),1:m,idx(:,1)')) = -1;
B(sub2ind(size(B),1:m,idx(:,2)')) = 1;
Approach #1
Here's a vectorized approach that has more memory requirements than a non-vectorized or for-loop based one. So, it could be tried out for small to medium sized datasizes.
The basic idea is this. For n=4 as an example, we take
-1 1 0 0
-1 0 1 0
-1 0 0 1
as the basic building block, replicate it n-1 i.e. 3 times and then remove the rows that aren't supposed to be part of the final output as per the requirements of the problem. Because of this very nature, this solution has more memory requirements, as we need to remove rows 6,8,9 for n = 4 case. But this gives us the opportunity to work with everything in one go.
N = n-1; %// minus 1 of the datasize, n
blksz = N*(N+1); %// number of elements in a (n-1)*n blocksize that is replicated
b1 = [-1*ones(N,1) eye(N)] %// Create that special starting (n-1)*n block
idx1 = find(b1~=0) %// find non zero elements for the starting block
idx2 = bsxfun(#plus,idx1,[0:N-1]*(blksz+N)) %// non zero elements for all blocks
b1nzr = repmat(b1(b1~=0),[1 N]) %// elements for all blocks
vald_ind = bsxfun(#le,idx2,[1:N]*blksz) %// positions of valid elements all blocks
mat1 = zeros(N,blksz) %// create an array for all blocks
mat1(idx2(vald_ind)) = b1nzr(vald_ind) %// put right elements into right places
%// reshape into a 3D array, join/concatenate along dim3
out = reshape(permute(reshape(mat1,N,N+1,[]),[1 3 2]),N*N,[])
%// remove rows that are not entertained according to the requirements of problem
out = out(any(out==1,2),:)
Approach #2
Here's a loop based code that could be easier to get a hold on if you have to explain it to yourself or just people and most importantly scales up pretty well on performance criteria across varying datasizes.
start_block = [-1*ones(n-1,1) eye(n-1)] %// Create that special starting (n-1)*n block
%// Find starting and ending row indices for each shifted block to be repeated
ends = cumsum([n-1:-1:1])
starts = [1 ends(1:end-1)+1]
out = zeros(sum(1:n-1),n) %// setup all zeros array to store output
for k1 = 1:n-1
%// Put elements from shifted portion of start_block for creating the output
out(starts(k1):ends(k1),k1:end) = start_block(1:n-k1,1:n-k1+1)
end
With n=4, the output -
out =
-1 1 0 0
-1 0 1 0
-1 0 0 1
0 -1 1 0
0 -1 0 1
0 0 -1 1
I don't know if I understood properly, but is this what you are looking for:
M=rand(5);
k=1; % this is to select the k-th diagonal
D=diag(ones(1,size(M,2)-abs(k)), k);
M(D==1)=-1;
M =
0.9834 -1.0000 0.8402 0.6310 0.0128
0.8963 0.1271 -1.0000 0.3164 0.6054
0.8657 0.6546 0.3788 -1.0000 0.5765
0.8010 0.8640 0.2682 0.4987 -1.0000
0.5550 0.2746 0.1529 0.7386 0.6550

Is there an array creation function for full arrays that has the same signature as sparse matrix constructor?

I'd like to accumulate indexed elements in a matrix, like table and tapply function in R.
I found sparse(i,j,s,m,n) fit my need perfectly,
As the document says:"Any elements of s that have duplicate values of i and j are added together."
But I have to convert the obtained sparse matrix to a full one using full():
a = a + full(sparse(i,j,s,m,n));
Is this a efficient way to do so?
By the way, is there anything like below, no matter whether adding duplicated i,j pairs?
a = setelements(a, i,j,s);
and
vector = getelement(a,i,j);
where i&j take meanings in sparse() function.
And what if a is a multidimensional array? sparse() only deal with matrix.
Do I have to set the entries page by page with outer loops?
Take a look at accumaray. For example,
ii = [1 2 2 3 3];
jj = [3 2 2 2 2];
s = [10 20 30 40 50];
a = accumarray([ii(:) jj(:)],s(:));
gives
a =
0 0 10
0 50 0
0 90 0
Note that each row of the first argument ([ii(:) jj(:)]) defines an N-dimensional index into the output array (N is 2 in this example).
accumarray is very flexible. It works for N-dimensional arrays, lets you specify size of the result (it may be larger than inferred from the supplied indices), and can even apply an arbitrary function (different from sum) to each set of values defined by the same index.
As a more general example, with the above data,
a = accumarray([ii(:) jj(:)],s(:),[4 4],#max)
gives
a =
0 0 10 0
0 30 0 0
0 50 0 0
0 0 0 0

Two FOR statements coupled into one

Is is possible to put two for statements into one statement. Something like
A = [ 0 0 0 5
0 2 0 0
1 3 0 0
0 0 4 0];
a=size(A);
b=size(A);
ind=0;
c=0;
for ({i=1:a},{j=1:b})
end
Your question is very broad, but one thing to consider is that in MATLAB you can often take advantage of linear indexing (instead of subscripting), without actually having to reshape the array. For example,
>> A = [ 0 0 0 5
0 2 0 0
1 3 0 0
0 0 4 0];
>> A(3,2)
ans =
3
>> A(7) % A(3+(2-1)*size(A,1))
ans =
3
You can often use this to your advantage in a for loop over all the elements:
for ii=1:numel(A),
A(ii) = A(ii) + 1; % or something more useful
end
Is the same as:
for ii=1:size(A,2),
for jj=1:size(A,1),
A(jj,ii) = A(jj,ii) + 1;
end
end
But to address your specific goal in this problem, as you stated in the comments ("I am storing the non zero elements in another matrix; with elements like the index number, value, row number and column number."), of making sparse matrix representation, it comes to this:
>> [i,j,s] = find(A);
>> [m,n] = size(A);
>> S = sparse(i,j,s,m,n)
S =
(3,1) 1
(2,2) 2
(3,2) 3
(4,3) 4
(1,4) 5
But that's not really relevant to the broader question.
Actually you can combine multiple loops into one for, however it would require you to loop over a vector containing all elements rather than the individual elements.
Here is a way to do it:
iRange = 1:2;
jRange = 1:3;
[iL jL] = ndgrid(iRange,jRange);
ijRange = [iL(:) jL(:)]';
for ij = ijRange
i = ij(1); j = ij(2);
end
Note that looping over the variables may be simpler, but perhaps this method has some advantages as well.
No
read this http://www.mathworks.com/help/matlab/matlab_prog/loop-control-statements.html
i also don't see any added value even if it was possible
No I don't think you can put two for loops in one line.
Depends on your operation, you may be able to reshape it and use one for loop. If you are doing something as simple as just printing out all elements,
B = reshape(A,a*b,1);
for i=1:a*b
c = B(i);
...
end