Related
I have large sets of 3D data consisting of 1D signals acquired in 2D space.
The first step in processing this data is thresholding all signals to find the arrival of a high-amplitude pulse. This pulse is present in all signals and arrives at different times.
After thresholding, the 3D data set should be reordered so that every signal starts at the arrival of the pulse and what came before is thrown away (the end of the signals is of no importance, as of now i concatenate zeros to the end of all signals so the data remains the same size).
Now, I have implemented this in the following manner:
First, i start by calculating the sample number of the first sample exceeding the threshold in all signals
M = randn(1000,500,500); % example matrix of realistic size
threshold = 0.25*max(M(:,1,1)); % 25% of the maximum in the first signal as threshold
[~,index] = max(M>threshold); % indices of first sample exceeding threshold in all signals
Next, I want all signals to be shifted so that they all start with the pulse. For now, I have implemented it this way:
outM = zeros(size(M)); % preallocation for speed
for i = 1:size(M,2)
for j = 1:size(M,3)
outM(1:size(M,1)+1-index(1,i,j),i,j) = M(index(1,i,j):end,i,j);
end
end
This works fine, and i know for-loops are not that slow anymore, but this easily takes a few seconds for the datasets on my machine. A single iteration of the for-loop takes about 0.05-0.1 sec, which seems slow to me for just copying a vector containing 500-2000 double values.
Therefore, I have looked into the best way to tackle this, but for now I haven't found anything better.
I have tried several things: 3D masks, linear indexing, and parallel loops (parfor).
for 3D masks, I checked to see if any improvements are possible. Therefore i first contruct a logical mask, and then compare the speed of the logical mask indexing/copying to the double nested for loop.
%% set up for logical mask copying
AA = logical(ones(500,1)); % only copy the first 500 values after the threshold value
Mask = logical(zeros(size(M)));
Jepla = zeros(500,size(M,2),size(M,3));
for i = 1:size(M,2)
for j = 1:size(M,3)
Mask(index(1,i,j):index(1,i,j)+499,i,j) = AA;
end
end
%% speed comparison
tic
Jepla = M(Mask);
toc
tic
for i = 1:size(M,2)
for j = 1:size(M,3)
outM(1:size(M,1)+1-index(1,i,j),i,j) = M(index(1,i,j):end,i,j);
end
end
toc
The for-loop is faster every time, even though there is more that's copied.
Next, linear indexing.
%% setup for linear index copying
%put all indices in 1 long column
LongIndex = reshape(index,numel(index),1);
% convert to linear indices and store in new variable
linearIndices = sub2ind(size(M),LongIndex,repmat(1:size(M,2),1,size(M,3))',repelem(1:size(M,3),size(M,2))');
% extend linear indices with those of all values to copy
k = zeros(numel(M),1);
count = 1;
for i = 1:numel(LongIndex)
values = linearIndices(i):size(M,1)*i;
k(count:count+length(values)-1) = values;
count = count + length(values);
end
k = k(1:count-1);
% get linear indices of locations in new matrix
l = zeros(length(k),1);
count = 1;
for i = 1:numel(LongIndex)
values = repelem(LongIndex(i)-1,size(M,1)-LongIndex(i)+1);
l(count:count+length(values)-1) = values;
count = count + length(values);
end
l = k-l;
% create new matrix
outM = zeros(size(M));
%% speed comparison
tic
outM(l) = M(k);
toc
tic
for i = 1:size(M,2)
for j = 1:size(M,3)
outM(1:size(M,1)+1-index(1,i,j),i,j) = M(index(1,i,j):end,i,j);
end
end
toc
Again, the alternative approach, linear indexing, is (a lot) slower.
After this failed, I learned about parallelisation, and though this would for sure speed up my code.
By reading some of the documentation around parfor and trying it out a bit, I changed my code to the following:
gcp;
outM = zeros(size(M));
inM = mat2cell(M,size(M,1),ones(size(M,2),1),size(M,3));
tic
parfor i = 1:500
for j = 1:500
outM(:,i,j) = [inM{i}(index(1,i,j):end,1,j);zeros(index(1,i,j)-1,1)];
end
end
end
toc
I changed it so that "outM" and "inM" would both be sliced variables, as I read this is best. Still this is very slow, a lot slower than the original for loop.
So now the question, should I give up on trying to improve the speed of this operation? Or is there another way in which to do this? I have searched a lot, and for now do not see how to speed this up.
Sorry for the long question, but I wanted to show what I tried.
Thank you in advance!
Not sure if an option in your situation, but looks like cell arrays are actually faster here:
outM2 = cell(size(M,2),size(M,3));
tic;
for i = 1:size(M,2)
for j = 1:size(M,3)
outM2{i,j} = M(index(1,i,j):end,i,j);
end
end
toc
And a second idea which also came out faster, batch all data which have to be shifted by the same value:
tic;
for i = 1:unique(index).'
outM(1:size(M,1)+1-i,index==i) = M(i:end,index==i);
end
toc
It totally depends on your data if this approach is actually faster.
And yes integer valued and logical indexing can be mixed
I have a 3D Mesh grid, X, Y, Z. I want to create a new 3D array that is a function of X, Y, & Z. That function comprises the sum of several 3D Gaussians located at different points. Currently, I have a for loop that runs over the different points where I have my gaussians, and I have an array of center locations r0(nGauss, 1:3)
[X,Y,Z]=meshgrid(-10:.1:10);
Psi=0*X;
for index = 1:nGauss
Psi = Psi + Gauss3D(X,Y,Z,[r0(index,1),r0(index,2),r0(index,3)]);
end
where my 3D gaussian function is
function output=Gauss3D(X,Y,Z,r0)
output=exp(-(X-r0(1)).^2 + (Y-r0(2)).^2 + (Z-r0(3)).^2);
end
I'm happy to redesign the function, which is the slowest part of my code and has to happen many many time, but I can't figure out how to vectorize this so that it will run faster. Any suggestions would be appreciated
*****NB the Original function had a square root in it, and has been modified to make it an actual gaussian***
NOTE! I've modified your code to create a Gaussian, which was:
output=exp(-sqrt((X-r0(1)).^2 + (Y-r0(2)).^2 + (Z-r0(3)).^2));
That does not make a Gaussian. I changed this to:
output = exp(-((X-r0(1)).^2 + (Y-r0(2)).^2 + (Z-r0(3)).^2));
(note no sqrt). This is a Gaussian with sigma = sqrt(1/2).
If this is not what you want, then this answer might not be very useful to you, because your function does not go to 0 as fast as the Gaussian, and therefore is harder to truncate, and it is not separable.
Vectorizing this code is pointless, as the other answers attest. MATLAB's JIT is perfectly capable of running this as fast as it'll go. But you can reduce the amount of computation significantly by noting that the Gaussian goes to almost zero very quickly, and is separable:
Most of the exp evaluations you're doing here yield a very tiny number. You don't need to compute those, just fill in 0.
exp(-x.^2-y.^2) is the same as exp(-x.^2).*exp(-y.^2), which is much cheaper to compute.
Let's put these two things to the test. Here is the test code:
function gaussian_test
N = 100;
r0 = rand(N,3)*20 - 10;
% Original
tic
[X,Y,Z] = meshgrid(-10:.1:10);
Psi1 = zeros(size(X));
for index = 1:N
Psi1 = Psi1 + Gauss3D(X,Y,Z,r0(index,:));
end
t = toc;
fprintf('original, time = %f\n',t)
% Fast, large truncation
tic
[X,Y,Z] = deal(-10:.1:10);
Psi2 = zeros(numel(X),numel(Y),numel(Z));
for index = 1:N
Psi2 = Gauss3D_fast(Psi2,X,Y,Z,r0(index,:),5);
end
t = toc;
fprintf('tuncation = 5, time = %f\n',t)
fprintf('mean abs error = %f\n',mean(reshape(abs(Psi2-Psi1),[],1)))
fprintf('mean square error = %f\n',mean(reshape((Psi2-Psi1).^2,[],1)))
fprintf('max abs error = %f\n',max(reshape(abs(Psi2-Psi1),[],1)))
% Fast, smaller truncation
tic
[X,Y,Z] = deal(-10:.1:10);
Psi3 = zeros(numel(X),numel(Y),numel(Z));
for index = 1:N
Psi3 = Gauss3D_fast(Psi3,X,Y,Z,r0(index,:),3);
end
t = toc;
fprintf('tuncation = 3, time = %f\n',t)
fprintf('mean abs error = %f\n',mean(reshape(abs(Psi3-Psi1),[],1)))
fprintf('mean square error = %f\n',mean(reshape((Psi3-Psi1).^2,[],1)))
fprintf('max abs error = %f\n',max(reshape(abs(Psi3-Psi1),[],1)))
% DIPimage, same smaller truncation
tic
Psi4 = newim(201,201,201);
coords = (r0+10) * 10;
Psi4 = gaussianblob(Psi4,coords,10*sqrt(1/2),(pi*100).^(3/2));
t = toc;
fprintf('DIPimage, time = %f\n',t)
fprintf('mean abs error = %f\n',mean(reshape(abs(Psi4-Psi1),[],1)))
fprintf('mean square error = %f\n',mean(reshape((Psi4-Psi1).^2,[],1)))
fprintf('max abs error = %f\n',max(reshape(abs(Psi4-Psi1),[],1)))
end % of function gaussian_test
function output = Gauss3D(X,Y,Z,r0)
output = exp(-((X-r0(1)).^2 + (Y-r0(2)).^2 + (Z-r0(3)).^2));
end
function Psi = Gauss3D_fast(Psi,X,Y,Z,r0,trunc)
% sigma = sqrt(1/2)
x = X-r0(1);
y = Y-r0(2);
z = Z-r0(3);
mx = abs(x) < trunc*sqrt(1/2);
my = abs(y) < trunc*sqrt(1/2);
mz = abs(z) < trunc*sqrt(1/2);
Psi(my,mx,mz) = Psi(my,mx,mz) + exp(-x(mx).^2) .* reshape(exp(-y(my).^2),[],1) .* reshape(exp(-z(mz).^2),1,1,[]);
% Note! the line above uses implicit singleton expansion. For older MATLABs use bsxfun
end
This is the output on my machine, reordered for readability (I'm still on MATLAB R2017a):
| time(s) | mean abs | mean sq. | max abs
--------------+----------+----------+----------+----------
original | 5.035762 | | |
tuncation = 5 | 0.169807 | 0.000000 | 0.000000 | 0.000005
tuncation = 3 | 0.054737 | 0.000452 | 0.000002 | 0.024378
DIPimage | 0.044099 | 0.000452 | 0.000002 | 0.024378
As you can see, using these two properties of the Gaussian we can reduce time from 5.0 s to 0.17 s, a 30x speedup, with hardly noticeable differences (truncating at 5*sigma). A further 3x speedup can be gained by allowing a small error. The smallest the truncation value, the faster this will go, but the larger the error will be.
I added that last method, the gaussianblob function from DIPimage (I'm an author), just to show that option in case you need to squeeze that bit of extra time from your code. That function is implemented in C++. This version that I used you will need to compile yourself. Our current official release implements this function still in M-file code, and is not as fast.
Further chance of improvement is if the fractional part of the coordinates is always the same (w.r.t. the pixel grid). In this case, you can draw the Gaussian once, and shift it over to each of the centroids.
Another alternative involves computing the Gaussian once, at a somewhat larger scale, and interpolating into it to generate each of the 1D Gaussians needed to generate the output. I did not implement this, I have no idea if it will be faster or if the time difference will be significant. In the old days, exp was expensive, I'm not sure this is still the case.
So, I am building off of the answer above me #Durkee. I enjoy these kinds of problems, so I thought a little about how to make each of the expansions implicit, and I have the one-line function below. Using this function I shaved .11 seconds off of the call, which is completely negligible. It looks like yours is pretty decent. The only advantage of mine might be how the code scales on a finer mesh.
xLin = [-10:.1:10]';
tic
psi2 = sum(exp(-sqrt((permute(xLin-r0(:,1)',[3 1 4 2])).^2 ...
+ (permute(xLin-r0(:,2)',[1 3 4 2])).^2 ...
+ (permute(xLin-r0(:,3)',[3 4 1 2])).^2)),4);
toc
The relative run times on my computer were (all things kept the same):
Original - 1.234085
Other - 2.445375
Mine - 1.120701
So this is a bit of an unusual problem where on my computer the unvectorized code actually works better than the vectorized code, here is my script
clear
[X,Y,Z]=meshgrid(-10:.1:10);
Psi=0*X;
nGauss = 20; %Sample nGauss as you didn't specify
r0 = rand(nGauss,3); % Just make this up as it doesn't really matter in this case
% Your original code
tic
for index = 1:nGauss
Psi = Psi + Gauss3D(X,Y,Z,[r0(index,1),r0(index,2),r0(index,3)]);
end
toc
% Vectorize these functions so we can use implicit broadcasting
X1 = X(:);
Y1 = Y(:);
Z1 = Z(:);
tic
val = [X1 Y1 Z1];
% Change the dimensions so that r0 operates on the right elements
r0_temp = permute(r0,[3 2 1]);
% Perform the gaussian combination
out = sum(exp(-sqrt(sum((val-r0_temp).^2,2))),3);
toc
% Check to make sure both functions match
sum(abs(vec(Psi)-vec(out)))
function output=Gauss3D(X,Y,Z,r0)
output=exp(-sqrt((X-r0(1)).^2 + (Y-r0(2)).^2 + (Z-r0(3)).^2));
end
function out = vec(in)
out = in(:);
end
As you can see, this is probably about as vectorized as you can get. The whole function is done using broadcasting and vectorized operations which normally improve performance ten-one hundredfold. However, in this case, this is not what we see
Elapsed time is 1.876460 seconds.
Elapsed time is 2.909152 seconds.
This actually shows the unvectorized version as being faster.
There could be a few reasons for this of which I am by no means an expert.
MATLAB uses a JIT compiler now which means that for loops are no longer inefficient.
Your code is already reasonably vectorized, you are operating at 8 million elements at once
Unless nGauss is 1000 or something, you're not looping through that much, and at that point, vectorization means you will run out of memory
I could be hitting some memory threshold where I am using too much memory and that is making my code inefficient, I noticed that when I lowered the resolution on the meshgrid the vectorized version worked better
As an aside, I tested this on my GTX 1060 GPU with single precision(single precision is 10x faster than double precision on most GPUs)
Elapsed time is 0.087405 seconds.
Elapsed time is 0.241456 seconds.
Once again the unvectorized version is faster, sorry I couldn't help you out but it seems that your code is about as good as you are going to get unless you lower the tolerances on your meshgrid.
I have a 10000x1 matrix. I need to find the percentage of information contained in each case. The method of doing so is to make another matrix that contains the sum of remaining cells.
Example:
% info is a 10000x1
% info_n contains percentages of information.
info_n = zeros([10000 1]);
info_n(1)= info(1) /sum(info);
info_n(2)=(info(1)+info(2)) /sum(info);
info_n(3)=(info(1)+info(2)+info(3))/sum(info);
I need to do this but all the way to info_n(10000).
So far, I tried this:
for i = 1:10000
info_n(i)=(info(i))/sum(info);
end
but I can't think of a way to add previous information. Thank you for the help!
You can use cumsum ("cumulative sum") for this task:
info_n = cumsum(info)/sum(info);
EDIT: #LuisMendo's comment sparked my curiosity. It seems that you actually gain something by using his method if the length N of the vector is below about 2^15 but the advantage after that drops. This is because cumsum probably needs O(N^2) but sum only needs O(N) time.
On the left you see the times as well as the ratio of the times of the two methods plotted against the logarithm of the size. On the right you see the actual absolute time differences, which varies a lot depending on what else is currently running:
Testing script:
N = 2.^(1:28);
y1 = zeros(size(N));
y2 = zeros(size(N));
vec = 1:numel(N);
for k=vec;
disp(k)
info = zeros(N(k),1);
% flawr's suggestion
tic
info_n = cumsum(info);
info_n = info_n / info_n(end);
y1(k) = toc;
% LuisMendo's suggestion
tic
info_m = cumsum(info)/sum(info);
y2(k) = toc;
end
subplot(1,2,1)
semilogy(vec,y1,vec,y2,vec,y1./y2,vec,vec.^0);
legend('LuisMendo','flawr','LM/f','Location','SouthEast');
subplot(1,2,2)
plot(vec,y1-y2)
I'm trying to estimate the (unknown) original datapoints that went into calculating a (known) moving average. However, I do know some of the original datapoints, and I'm not sure how to use that information.
I am using the method given in the answers here: https://stats.stackexchange.com/questions/67907/extract-data-points-from-moving-average, but in MATLAB (my code below). This method works quite well for large numbers of data points (>1000), but less well with fewer data points, as you'd expect.
window = 3;
datapoints = 150;
data = 3*rand(1,datapoints)+50;
moving_averages = [];
for i = window:size(data,2)
moving_averages(i) = mean(data(i+1-window:i));
end
length = size(moving_averages,2)+(window-1);
a = (tril(ones(length,length),window-1) - tril(ones(length,length),-1))/window;
a = a(1:length-(window-1),:);
ai = pinv(a);
daily = mtimes(ai,moving_averages');
x = 1:size(data,2);
figure(1)
hold on
plot(x,data,'Color','b');
plot(x(window:end),moving_averages(window:end),'Linewidth',2,'Color','r');
plot(x,daily(window:end),'Color','g');
hold off
axis([0 size(x,2) min(daily(window:end))-1 max(daily(window:end))+1])
legend('original data','moving average','back-calculated')
Now, say I know a smattering of the original data points. I'm having trouble figuring how might I use that information to more accurately calculate the rest. Thank you for any assistance.
You should be able to calculate the original data exactly if you at any time can exactly determine one window's worth of data, i.e. in this case n-1 samples in a window of length n. (In your case) if you know A,B and (A+B+C)/3, you can solve now and know C. Now when you have (B+C+D)/3 (your moving average) you can exactly solve for D. Rinse and repeat. This logic works going backwards too.
Here is an example with the same idea:
% the actual vector of values
a = cumsum(rand(150,1) - 0.5);
% compute moving average
win = 3; % sliding window length
idx = hankel(1:win, win:numel(a));
m = mean(a(idx));
% coefficient matrix: m(i) = sum(a(i:i+win-1))/win
A = repmat([ones(1,win) zeros(1,numel(a)-win)], numel(a)-win+1, 1);
for i=2:size(A,1)
A(i,:) = circshift(A(i-1,:), [0 1]);
end
A = A / win;
% solve linear system
%x = A \ m(:);
x = pinv(A) * m(:);
% plot and compare
subplot(211), plot(1:numel(a),a, 1:numel(m),m)
legend({'original','moving average'})
title(sprintf('length = %d, window = %d',numel(a),win))
subplot(212), plot(1:numel(a),a, 1:numel(a),x)
legend({'original','reconstructed'})
title(sprintf('error = %f',norm(x(:)-a(:))))
You can see the reconstruction error is very small, even using the data sizes in your example (150 samples with a 3-samples moving average).
Not quite sure what this means.
"Warning: Matrix is singular to working precision."
I have a 3x4 matrix called matrix bestM
matrix Q is 3x3 of bestM and matrix m is the last column of bestM
I would like to do C = -Inverse matrix of Q * matrix m
and I get that warning
and C =[Inf Inf Inf] which isn't right because i am calculating for the camera center in the world
bestM = [-0.0031 -0.0002 0.0005 0.9788;
-0.0003 -0.0006 0.0028 0.2047;
-0.0000 -0.0000 0.0000 0.0013];
Q = bestM(1:3,1:3);
m = bestM(:,4);
X = inv(Q);
C = -X*m;
disp(C);
A singular matrix can be thought of as the matrix equivalent of zero, when you try to invert 0 it blows up (goes to infinity) which is what you are getting here. user 1281385 is absolutely wrong about using the format command to increase precision; the format command is used to change the format of what is shown to you. In fact the very first line of the help command for format says
format does not affect how MATLAB computations are done.
As found here, a singular matrix is one that does not have an inverse. As dvreed77 already pointed out, you can think of this as 1/0 for matrices.
Why I'm answering, is to tell you that using inv explicitly is almost never a good idea. If you need the same inverse a few hundred times, it might be worth it, however, in most circumstances you're interested in the product C:
C = -inv(Q)*m
which can be computed much more accurately and faster in Matlab using the backslash operator:
C = -Q\m
Type help slash for more information on that. And even if you happen to find yourself in a situation where you really need the inverse explicitly, I'd still advise you to avoid inv:
invQ = Q\eye(size(Q))
Below is a little performance test to demonstrate one of the very few situations where the explicit inverse can be handy:
% This test will demonstrate the one case I ever encountered where
% an explicit inverse proved useful. Unfortunately, I cannot disclose
% the full details without breaking the law, but roughly, it came down
% to this: The (large) design matrix A, a result of a few hundred
% co-registrated images, needed to be used to solve several thousands
% of systems, where the result matrices b came from processing the
% images one-by-one.
%
% That means the same design matrix was re-used thousands of times, to
% solve thousands of systems at a time. To add to the fun, the images
% were also complex-valued, but I'll leave that one out of consideration
% for now :)
clear; clc
% parameters for this demo
its = 1e2;
sz = 2e3;
Bsz = 2e2;
% initialize design matrix
A = rand(sz);
% initialize cell-array to prevent allocating memory from consuming
% unfair amounts of time in the first loop.
% Also, initialize them, NOT copy them (as in D=C,E=D), because Matlab
% follows a lazy copy-on-write scheme, which would influence the results
C = {cellfun(#(~) zeros(sz,Bsz), cell(its,1), 'uni', false) zeros(its,1)};
D = {cellfun(#(~) zeros(sz,Bsz), cell(its,1), 'uni', false) zeros(its,1)};
E = {cellfun(#(~) zeros(sz,Bsz), cell(its,1), 'uni', false) zeros(its,1)};
% The impact of rand() is the same in both loops, so it has no
% effect, it just gives a longer total run time. Still, we do the
% rand explicitly to *include* the indexing operation in the test.
% Also, caching will most definitely influence the results, because
% any compiler (JIT), even without optimizations, might recognize the
% easy performance gain when the code computes the same array over and
% over again. It probably will, but we have no control over when and
% wherethat happens. So, we prevent that from happening at all, by
% re-initializing b at every iteration.
% The assignment to cell is a necessary part of the demonstration;
% it is the desired output of the whole calculation. Assigning to cell
% instead of overwriting 'ans' takes some time, which is to be included
% in the demonstration, again for cache reasons: the extra time is now
% guaranteed to be equal in both loops, so it really does not matter --
% only the total run time will be affected.
% Direct computation
start = tic;
for ii = 1:its
b = rand(sz,Bsz);
C{ii,1} = A\b;
C{ii,2} = max(max(abs( A*C{ii,1}-b )));
end
time0 = toc(start);
[max([C{:,2}]) mean([C{:,2}]) std([C{:,2}])]
% LU factorization (everyone's
start = tic;
[L,U,P] = lu(A, 'vector');
for ii = 1:its
b = rand(sz,Bsz);
D{ii,1} = U\(L\b(P,:));
D{ii,2} = max(max(abs( A*D{ii,1}-b )));
end
time1 = toc(start);
[max([D{:,2}]) mean([D{:,2}]) std([D{:,2}])]
% explicit inv
start = tic;
invA = A\eye(size(A)); % NOTE: DON'T EVER USE INV()!
for ii = 1:its
b = rand(sz,Bsz);
E{ii,1} = invA*b;
E{ii,2} = max(max(abs( A*E{ii,1}-b )));
end
time2 = toc(start);
[max([E{:,2}]) mean([E{:,2}]) std([E{:,2}])]
speedup0_1 = (time0/time1-1)*100
speedup1_2 = (time1/time2-1)*100
speedup0_2 = (time0/time2-1)*100
Results:
% |Ax-b|
1.0e-12 * % max. mean st.dev.
0.1121 0.0764 0.0159 % A\b
0.1167 0.0784 0.0183 % U\(L\b(P,;))
0.0968 0.0845 0.0078 % invA*b
speedup0_1 = 352.57 % percent
speedup1_2 = 12.86 % percent
speedup0_2 = 410.80 % percent
It should be clear that an explicit inverse has its uses, but just as a goto construct in any language -- use it sparingly and wisely.