how can i get the cumsum of a 3D matrix.

i have a 3-D matrix.I want to use the cumsum algorithm to plot all the matrix as a signal. i have the following code

m=zeros(10,10,30);  %define 3-D 10 by 10 by 30 matrix
n=30;               %number of matrix

for k=1:n
    t=linspace(0,1.0,n);    %specify the value an element will take in each iteration
    m(3,3,k) = t(k);    %the 3rd row and 3rd column in each matrix will have the value of t
    disp(m(:,:,k))
end
for k=1:n
    a=cusum(m(:,:,k))
    plot (a)
    hold on
end

[Geoff Hayes]

Folakemi - what exactly are you trying to plot from that matrix? Won't a be a single value if you use cumsum (not the cusum that your code is referencing)?

[Folakemi Omotoye]

from the code, only one pixel is changing in the matrix. i need to plot all the matrix as a signal and specify a threshold, such that the signal will reflect the threshold and when changing pixel exceeds the threshold

[Geoff Hayes]

What do you mean by only one pixel is changing in the matrix? Do you mean across dimensions? The m(3,3,k) for all k will be the same matrix since t never changes....

[Folakemi Omotoye]

only the 3rd row 3rd column will be changing (line 5 of the code)

[Geoff Hayes]

Since your m(:,:,k) is a 10x10 matrix, then what output are you expecting when you use cumsum? Do you want another 10x10 matrix or do you want a 10x1 or a scalar?

[Folakemi Omotoye]

something like that, but how will i be able to plot a signal with a 10 x 10 matrix

[Geoff Hayes]

something like that..well what exactly do you want to plot?

[Folakemi Omotoye]

plot the changes in all the matrices as a signal

[Rik]

Could you describe your intended resulting plot by describing how the the x and y values of each point in the plot should result from your matrix?

[Folakemi Omotoye]

The aim is to achieve a signal in this form (attached picture)

Answer 1

[Constantino Carlos Reyes-Aldasoro]

To be able to properly help you, the problem needs to be well defined. As context, what you obtain with cumsum is an accummulation of values along one dimension of the data. So you can find this accummulation along any dimension you want. Think of it with a 1D matrix, many values and they accummulate within the dimension. If you have a probability distribution, the last value of the cumsum should always be 1.

If what you want is to find a projection, or in proper statistical terms, a marginal distribution, then you need to use sum( , dim) along the dimension of interest.

[Folakemi Omotoye]

I tried the cumsum and i specified the dimension but there was error saying dimensions must be the same. the aim is to achieve a signal in this form(see pic attached) of the changing pixels