I have a 3D matrix d and I want to find the signed minimum value along the third dimension. Currently, I use following code
tmp = abs(d);
[row, col]=ndgrid(1:size(d,1),1:size(d,2));
[v,ind] = min(tmp,[],3);
index = row + size(d,1)*size(d,2)*(ind-1)+ size(d,1)*(col-1); %turn the ind to index
dm = d(index); %get the signed minimum value
The above code does not so efficient. Does anyone know a better choice? Thank you!
Rather than creating your grid of row and col values to convert ind into an index which you can use to index back into d, you can just use the first output of min which contains those minimum values along the third dimension.
dm = min(abs(d), [], 3);
Related
I am implementing a neural network and am trying to one hot encode a matrix of column vectors based on the max value in each column. Previously, I had been iterating through the matrix vector by vector, but I've been told that this is unnecessary and that I can actually one hot encode every column vector in the matrix at the same time. Unfortunately, after perusing SO, GitHub, and MathWorks, nothing seems to be getting the job done. I've listed my previous code below. Please help! Thanks :)
UPDATE:
This is what I am trying to accomplish...except this only changed the max value in the entire matrix to 1. I want to change the max value in each COLUMN to 1.
one_hots = bsxfun(#eq, mini_batch_activations, max(mini_batch_activations(:)))
UPDATE 2:
This is what I am looking for, but it only works for rows. I need columns.
V = max(mini_batch_activations,[],2);
idx = mini_batch_activations == V;
Iterative code:
% This is the matrix I want to one hot encode
mini_batch_activations = activations{length(layers)};
%For each vector in the mini_batch:
for m = 1:size(mini_batch_activations, 2)
% Isolate column vector for mini_batch
vector = mini_batch_activations(:,m);
% One hot encode vector to compare to target vector
one_hot = zeros(size(mini_batch_activations, 1),1);
[max_val,ind] = max(vector);
one_hot(ind) = 1;
% Isolate corresponding column vector in targets
mini_batch = mini_batch_y{k};
target_vector = mini_batch(:,m);
% Compare one_hot to target vector , and increment result if they match
if isequal(one_hot, target_vector)
num_correct = num_correct + 1;
endif
...
endfor
You’ve got the maxima for each column:
V = max(mini_batch_activations,[],1); % note 1, not 2!
Now all you need to do is equality comparison, the output is a logical array that readily converts to 0s and 1s. Note that MATLAB and Octave do implicit singleton expansion:
one_hot = mini_batch_activations==V;
I have two matrices A and B. A(:,1) corresponds to an x-coordinate, A(:,2) corresponds to a y-coordinate, and A(:,3) corresponds to a certain radius. All three values in a row describe the same circle. Now let's say...
A =
[1,4,3]
[8,8,7]
[3,6,3]
B =
[1,3,3]
[1, 92,3]
[4,57,8]
[5,62,1]
[3,4,6]
[9,8,7]
What I need is to be able to loop through matrix A and determine if there are any rows in matrix B that are "similar" as in the x value is within a range (-2,2) of the x value of A (Likewise with the y-coordinate and radius).If it satisfies all three of these conditions, it will be added to a new matrix with the values that were in A. So for example I would need the above data to return...
ans =
[1,4,3]
[8,8,7]
Please help and thank you in advance to anyone willing to take the time!
You can use ismembertol.
result = A(ismembertol(A,B,2,'ByRows',1,'DataScale',1),:)
Manual method
A = [1,4,3;
8,8,7;
3,6,3];
B = [1,3,3;
1,92,3;
4,57,8;
5,62,1;
3,4,6;
9,8,7]; % example matrices
t = 2; % desired threshold
m = any(all(abs(bsxfun(#minus, A, permute(B, [3 2 1])))<=t, 2), 3);
result = A(m,:);
The key is using permute to move the first dimension of B to the third dimension. Then bsxfun computes the element-wise differences for all pairs of rows in the original matrices. A row of A should be selected if all the absolute differences with respect to any column of B are less than the desired threshold t. The resulting variable m is a logical index which is used for selecting those rows.
Using pdist2 (Statistics and Machine Learning Toolbox)
m = any(pdist2(A, B, 'chebychev')<=t, 2);
result = A(m,:);
Ths pdist2 function with the chebychev option computes the maximum coordinate difference (Chebychev distance, or L∞ metric) between pairs of rows.
With for loop
It should work:
A = [1,4,3;
8,8,7;
3,6,3]
B = [1,3,3;
1,92,3;
4,57,8;
5,62,1;
3,4,6;
9,8,7]
index = 1;
for i = 1:size(A,1)
C = abs(B - A(i,:));
if any(max(C,[],2)<=2)
out(index,:) = A(i,:);
index = index + 1
end
end
For each row of A, computes the absolute difference between B and that row, then checks if there exists a row in which the maximum is less than 2.
Without for loop
ind = any(max(abs(B - permute(A,[3 2 1])),[],2)<=2);
out = A(ind(:),:);
I have sum of 3 cell arrays
A=72x1
B=72x720
C=72x90
resultant=A+B+C
size of resultant=72x64800
now when I find the minimum value with row and column indices I can locate the row element easily but how can I locate the column element in variables?
for example
after dong calculations for A,B,C I added them all and got a resultant in from of <72x(720x90)> or can say a matrix of integers of size <72x64800> then I found the minimum value of resultant with row and column index using the code below.
[minimumValue,ind]=min(resultant(:));
[row,col]=find(result== minimumValue);
then row got 14 and column got 6840 value..
now I can trace row 14 of all A,B,C variables easily but how can I know that the resultant column 6480 belongs to which combination of A,B,C?
Instead of using find, use the ind output from the min function. This is the linear index for minimumValue. To do that you can use ind2sub:
[r,c] = ind2sub(size(resultant),ind);
It is not quite clear what do you mean by resultant = A+B+C since you clearly don't sum them if you get a bigger array (72x64800), on the other hand, this is not a simple concatenation ([A B C]) since this would result in a 72x811 array.
However, assuming this is a concatenation you can do the following:
% get the 2nd dimension size of all matrices:
cols = cellfun(#(x) size(x,2),{A,B,C})
% create a vector with reapiting matrices names for all their columns:
mats = repelem(['A' 'B' 'C'],cols);
% get the relevant matrix for the c column:
mats(c)
so mats(c) will be the matrix with the minimum value.
EDIT:
From your comment I understand that your code looks something like this:
% arbitrary data:
A = rand(72,1);
B = rand(72,720);
C = rand(72,90);
% initializing:
K = size(B,2);
N = size(C,2);
counter = 1;
resultant = zeros(72,K*N);
% summing:
for k = 1:K
for n = 1:N
resultant(:,counter) = A + B(:,k) + C(:,n);
counter = counter+1;
end
end
% finding the minimum value:
[minimumValue,ind] = min(resultant(:))
and from the start of the answer you know that you can do this:
[r,c] = ind2sub(size(resultant),ind)
to get the row and column of minimumValue in resultant. So, in the same way you can do:
[Ccol,Bcol] = ind2sub([N,K],c)
where Bcol and Ccol is the column in B and C, respectively, so that:
minimumValue == A(r) + B(r,Bcol) + C(r,Ccol)
To see how it's working imagine that the loop above fills a matrix M with the value of counter, and M has a size of N-by-K. Because we fill M with a linear index, it will be filled in a column-major way, so the row will correspond to the n iterator, and the column will correspond to the k iterator. Now c corresponds to the counter where we got the minimum value, and the row and column of counter in M tells us the columns in B and C, so we can use ind2sub again to get the subscripts of the position of counter. Off course, we don't really need to create M, because the values within it are just the linear indices themselves.
I have a huge matrix MxN matrix, say, A=rand([M,N]); and an index vector with N integer values between 1 and M, say, RandomIndex = randi(M,[1,N]);.
Now I would like to generate a row vector with entries
result = [A(RandomIndex(1),1), A(RandomIndex(2),2), ..., A(RandomIndex(N),N)]
What would be an efficient way to do this? It should be a very cheap operation but all my implementations are slow. I don't think there is a notation in Matlab to do this directly, is there?
The fastest option so far is
indexFunction = #(r,c) A(r,c);
result = cell2mat(arrayfun(indexFunction,RandomIndex,1:N,'UniformOutput',false));
Is there a more efficient way?
Use sub2ind
A(sub2ind(size(A), RandomIndex, 1:N))
sub2ind will convert the row and column indices given by RandomIndex and 1:N to linear indices based on size(A) which you can then use to index A directly.
Another way to do this is to use RandomIndex and 1:N to return an NxN matrix and then take the diagonal of this with diag
diag(A(RandomIndex, 1:N)).'
Note: .' is used to convert the row vector returned by diag to a column vector.
M=10;N=50;
A=rand([M,N]);
RandomIndex = randi(M,[1,N]);
out = zeros(1,numel(RandomIndex));
for ii = 1:numel(RandomIndex)
out(ii)=A(RandomIndex(ii),ii);
end
Another approach would be to use sparse and logical indexing:
M = sparse(RandomIndex, 1:N, 1) == 1;
out = A(M);
The first line of code generates a logical matrix where there is only 1 true value set in each column. This is defined by each value of RandomIndex. We convert this to a logical matrix, then index into your matrix to obtain the final random vector.
Use your index directly.
M = 100;N=100;
A = rand(M,N);
% get a random index that can be as large as your matrix
% 10 rows by 1 column
idx = randi(numel(A), 10,1);
t = A(idx);
I was trying to resample (with replacement) my database using 'bootstrap' in Matlab as follows:
D = load('Data.txt');
lead = D(:,1);
depth = D(:,2);
X = D(:,3);
Y = D(:,4);
%Bootstraping to resample 100 times
[resampling100,bootsam] = bootstrp(100,'corr',lead,depth);
%plottig the bootstraping result as histogram
hist(resampling100,10);
... ... ...
... ... ...
Though the script written above is correct, I wonder how I would be able to see/load the resampled 100 datasets created through bootstrap? 'bootsam(:)' display the indices of the data/values selected for the bootstrap samples, but not the new sample values!! Isn't it funny that I'm creating fake data from my original data and I can't even see what is created behind the scene?!?
My second question: is it possible to resample the whole matrix (in this case, D) altogether without using any function? However, I know how to create random values from a vector data using 'unidrnd'.
Thanks in advance for your help.
The answer to question 1 is that bootsam provides the indices of the resampled data. Specifically, the nth column of bootsam provides the indices of the nth resampled dataset. In your case, to obtain the nth resampled dataset you would use:
lead_resample_n = lead(bootsam(:, n));
depth_resample_n = depth(bootsam(:, n));
Regarding the second question, I'm guessing what you mean is, how would you just get a re-sampled dataset without worrying about applying a function to the resampled data. Personally, I would use randi, but in this situation, it is irrelevant whether you use randi or unidrnd. An example follows that assumes 4 columns of some data matrix D (as in your question):
%# Build an example dataset
T = 10;
D = randn(T, 4);
%# Obtain a set of random indices, ie indices of draws with replacement
Ind = randi(T, T, 1);
%# Obtain the resampled data
DResampled = D(Ind, :);
To create multiple re-sampled data, you can simply loop over the creation of random indices. Or you could do it in one step by creating a matrix of random indices and using that to index D. With careful use of reshape and permute you can turn this into a T*4*M array, where indexing m = 1, ..., M along the third dimension yields the mth resampled dataset. Example code follows:
%# Build an example dataset
T = 10;
M = 3;
D = randn(T, 4);
%# Obtain a set of random indices, ie indices of draws with replacement
Ind = randi(T, T, M);
%# Obtain the resampled data
DResampled = permute(reshape(D(Ind, :)', 4, T, []), [2 1 3]);