I am using Matlab and have 2 matrices:
A =
0 0 1
1 0 1
1 1 0
and
B =
0 0 1.1
1.0 0 0.8
1.2 0.8 0
My goal is to compare the 2 matrices and get the highest values of each i,j element and store such results into a third matrix C.
I am able to achieve such result by applying a for loop that checks every single element and then stores it into the matrix but i would like to have a more efficient and elegant way to do so. Can you suggest me some better way?
The result should be like:
C =
0 0 1.1
1 0 1
1.2 1 0
C = max(A,B)
Source on documentation
Related
I have an algorith that the number of possibles combinations of 0 and 1, can reach the number 2^39. Let's say i have n=2 situations, or n1=2^2=4 combinations of 0 and 1: 00,01,10,11.From that i can create an array a=zeros(n,n1) and fill the columns with the possible combinations? That means first column has 00,second 01,third 10,last 11.I want this to be dynamic that means that n can be 1,2,3...,39, show the array will be a=zeros(n,2^n).Thanks for any response!
Just for general understanding: why do you think you need an array of all combinations of all integers from 0 to 2³⁹? That array would consume 39×2³⁹/1000⁴ ≈ 21TB of RAM...last time I checked, only the world's most advanced supercomputers have such resources, and most people working with those machines consider generating arrays like this quite wasteful...
Anyway, for completeness, for any N, this is the simplest solution:
P = dec2bin(0:(2^N)-1)-'0'
But, a little piece of advice: dec2bin outputs character arrays. If you want numerical arrays, you can subtract the character '0', however, that gives you an array of doubles according to the rules of MATLAB:
>> P = dec2bin(0:(2^3)-1)-'0';
>> whos P
Name Size Bytes Class Attributes
P 8x3 192 double
If you want to minimize your memory consumption, generate a logical array instead:
>> P = dec2bin(0:(2^3)-1)=='1';
>> whos P
Name Size Bytes Class Attributes
P 8x3 24 logical
If you want to also speed up the execution, use the standard algorithm directly:
%// if you like cryptic one-liners
B1 = rem(floor((0:pow2(N)-1).' * pow2(1-N:0)), 2) == 1;
%// If you like readability
B = false(N,pow2(N));
V = 0:pow2(N)-1;
for ii = 1:N
B(ii,:) = rem(V,2)==1;
V = (V-B(ii,:))/2;
end
That last one (the loop) is fastest of all solutions for any N (at least on R2010b and R2013a), and it has the smallest peak memory (only 1/Nth of the cryptic one-liner).
So I'd go for that one :)
But, that's just me.
Using ndgrid with a comma-separated list as output (see also here):
[c{1:N}] = ndgrid(logical([0 1]));
c = cat(N+1,c{N:-1:1});
c = reshape(c,[],N);
Example: N=4 gives
c =
0 0 0 0
0 0 0 1
0 0 1 0
0 0 1 1
0 1 0 0
0 1 0 1
0 1 1 0
0 1 1 1
1 0 0 0
1 0 0 1
1 0 1 0
1 0 1 1
1 1 0 0
1 1 0 1
1 1 1 0
1 1 1 1
I would like to enter the same vector of numbers repeatedly to an existing matrix at specific (row) logical indices. This is like an extension of entering just a single number at all logical index positions (at least in my head).
I.e., it is possible to have
mat = zeros(5,3);
rowInd = logical([0 1 0 0 1]); %normally obtained from previous operation
mat(rowInd,1) = 15;
mat =
0 0 0
15 0 0
0 0 0
0 0 0
15 0 0
But I would like to do sth like this
mat(rowInd,:) = [15 6 3]; %rows 2 and 5 should be filled with these numbers
and get an assignment mismatch error.
I want to avoid for loops for the rows or assigning vector elements single file. I have the strong feeling there is an elementary matlab operation that should be able to do this? Thanks!
The problem is that your indexing picks two rows from the matrix and tries to assign a single row to them. You have to replicate the targeted row to fit your indexing:
mat = zeros(5,3);
rowInd = logical([0 1 0 0 1]);
mat(rowInd,:) = repmat([15 6 3],sum(rowInd),1)
This returns:
mat =
0 0 0
15 6 3
0 0 0
0 0 0
15 6 3
I have a binary sparse matrix H of size 600*1200 constructed by concatenating square permutation matrices of size 200, thus sparse matrix have 3 ones in each column and 6 ones in each row. Now am trying to transform the matrix into reduced echelon form.
This is my code:
[m,n]=size(H);
for i=1:m
ind=find(H(:,i),1,'last');
if ind<=i
continue;
end
if ind~=i
temp=H(ind,:);
H(ind,:)=H(i,:);
H(i,:)=temp;
end
I=find(H(:,i));
% Guassian elimination
for j=1:length(I)
if I(j)~=i
H(I(j),:)=mod(H(I(j),:)+H(i,:),2);
end
end
end
But whichever H matrix generated, I can't get rid of other entries at 400th column,
how can I fix this, help
Since Gaussian elimination does not involve rearrangement of columns, the first 600 columns of the resulting matrix will only form the identity matrix if the first 600 columns of the original matrix were linearly independent. Otherwise, you are going to have "short" columns with other entries in them, as the Wikipedia article shows.
The way your matrix is structured, the first 400 columns are guaranteed to be linearly dependent. Indeed, the sum of the first 200 columns is the all-1 vector, and so is the sum of the columns 201-400. This is why you are seeing these entries in the 400th column.
To create identity matrix on the left you need to rearrange columns. One way, which looks a bit redundant but is very easy to code, is to
run your algorithm
rearrange columns (it's easy to identify the desired columns after rref)
do rref again.
Here is the code that does steps 2-3.
for i = 1:m
j = find(H(i,:),1,'first');
[H(:,i), H(:,j)] = deal(H(:,j), H(:,i));
end
H = rref(H)
Sample input into rref:
1 0 1 0 1
0 1 0 0 1
1 0 1 1 0
0 0 0 0 1
Output of your code:
1 0 1 1 0
0 1 0 0 1
0 0 0 1 1
0 0 0 0 1
After the column swap and second rref:
1 0 0 0 1
0 1 0 0 0
0 0 1 0 0
0 0 0 1 0
I would line to find the number of the first consecutive zero elements. For example in [0 0 1 -5 3 0] we have two zero consecutive elements that appear first in the vector.
could you please suggest a way without using for loops?
V=[0 0 1 -5 3 0] ;
k=find(V);
Number_of_first_zeros=k(1)-1;
Or,
Number_of_first_zeros=find(V,1,'first')-1;
To solve #The minion comment (if that was the purpose):
Number_of_first_zeros=find(V(find(~V,1,'first'):end),1,'first')-find(~V,1,'first');
Use a logical array to find the zeros and then look at where the zeros and ones are alternating.
V=[1 2 0 0 0 3 5123];
diff(V==0)
ans =
0 1 0 0 -1 0
Create sample data
V=[1 2 0 0 0 3 5123];
Find the zeros. The result will be a logical array where 1 represents the location of the zeros
D=V==0
D =
0 0 1 1 1 0 0
Take the difference of that array. 1 would then represent the start and -1 would represent the end.
T= diff(D)
ans =
0 1 0 0 -1 0
find(T==1) would give you the start and find(T==-1) would give you the end. The first index+1 of T==1 would be the start of the first set of zeros and the first index of T==-1 would be the end of the first set of zeros.
You could find position the first nonzero element using find.
I=find(A, 1);
The number of leading zeros is then I-1.
My solution is quite complex yet it doesn't use the loops and it does the trick. I am pretty sure, that there is a more direct approach.
Just in case no one else posts a working solution here my idea.
x=[1 2 4 0 20 0 10 1 23 45];
x1=find(x==0);
if numel(x1)>1
x2=[x1(2:end), 0];
x3=x2-x1;
y=find(x3~=1);
y(1)
elseif numel(x1)==1
display(1)
else
display('No zero found')
end
x is the dataset. x1 contains the index of all zero elements. x2 contains all those indices except the first one (because matrix dimensions must agree, one zero is added. x3 is the difference between the index and the previous index of zeros in your dataset. Now I find all those differences which are not 1 (do not correspond to sequences of zeros) and the first index (of this data is the required result. The if case is needed in case you have only one or no zero at all.
I'm assuming your question is the following: for the following vector [0 0 1 -5 3 0], I would like to find the index of the first element of a pair of 0 values. Is this correct? Therefore, the desired output for your vector would be '1'?
To extend the other answers to find any such pairs, not just 0 0 (eg. 0 1, 0 2, 3 4 etc), then this might help.
% define the pattern
ptrn = [ 0 0 ];
difference = ptrn(2) - ptrn(1)
V = [0 0 1 -5 3 0 0 2 3 4 0 0 1 0 0 0]
x = diff(V) == difference
indices = find(x)
indices =
1 6 11 14 15
This question already has answers here:
Construct this matrix based on two vectors MATLAB
(3 answers)
Closed 8 years ago.
I have a vector y = [0; 2; 4]
I want to convert each element of it into vector, where all elements are zero but element with index equal to digit is 1.
I'd like to do it without loops.
For example [0; 2; 4] should be converted to
[1 0 0 0 0 0 0 0 0 0;
0 0 1 0 0 0 0 0 0 0;
0 0 0 0 1 0 0 0 0 0]
(in this example vector first index is 0)
The usual trick with sparse can be used to simplify the process. Let n denote the desired number of columns. Then
result = full(sparse(1:numel(y), y+1, 1, numel(y), n));
For example, y = [0;2;4] and 10 produce
result =
1 0 0 0 0 0 0 0 0 0
0 0 1 0 0 0 0 0 0 0
0 0 0 0 1 0 0 0 0 0
First you need to decide how many digits you want to represent each number. In your case, you have 10 digits per number, so let's keep that in mind.
Once you do this, it's just a matter of indexing each element in your matrix. In your case, you have 10 digits per number. As such, do something like this:
y = [0; 2; 4]; %// Your digits array
out = zeros(numel(y), 10); %// 10 digits per number
ind = sub2ind(size(out), [1:numel(y)].', y+1);
out(ind) = 1;
The output should look like this:
out =
1 0 0 0 0 0 0 0 0 0
0 0 1 0 0 0 0 0 0 0
0 0 0 0 1 0 0 0 0 0
Let's go through this code slowly. y defines the digits you want per row of the output matrix. out allocates a matrix of zeroes where the number of rows is defined by how many digits you want in y. out will thus store your resulting matrix that you have shown us in your post.
The number of columns is 10, but you change this to be whatever you want. ind uses a command called sub2ind. This allows to completely vectorize the assignment of values in your out matrix and avoids a for loop. The first parameter is an array of values that defines how many rows and columns are in your matrix that you are trying to assign things to. In this case, it's just the size of out. The second and third parameters are the rows and columns you want to access in your matrix. In this case, the rows vary from 1 to as many elements as there are in y. In our case, this is 3. We want to generate one number per row, which is why it goes from 1 to 3. The columns denote where we want to set the digit to one for each row. As MATLAB indexes starting at 1, we have to make sure that we take y and add by 1. ind thus creates the column-major indices in order to access our matrix. The last statement finally accesses these locations and assigns a 1 to each location, thus producing our matrix.
Hope this helps!