Create a 4D matrix with specific range values

I would like to create a 4D matrix with specific range values:

For the first dimension, values from [5 10 15 ... 200] so basically 5*[1:40]

For the second dimension, values from the list [8 10 20 25 40 50 100 200]

For the third dimension, values from [1 2 3 ... 10] so [1:10]

For the fourth dimension, values from [1 2 ... 50] that is [1:50]

When accessing a specific "value" of that matrix, I would like to be able to view these rages that I've set up for each dimension. For example: A(2,1,2,3) = [10 8 2 3] (in this case)

How can I do that?

Thank you.

[James Tursa]

I don't really understand what it is you are trying to build. In particular, for a 4D matrix the expression A(2,1,2,3) would yield a scalar, not a vector as you have shown. How is this going to be used downstream in your code? Maybe we can suggest an alternate approach.

[Bruno Silva]

Hi James, thank you for your input. I understand what you said, I'll try to explain my point of view: let's say that, on A(2,1,2,3) I would like to store the vector [10;8;2;3]. Did I explain myself better? Thank you.

[Image Analyst]

No. You can't have a vector of 4 numbers at a particular location. You can just have a single number as James said. And you didn't even explain it - you just repeated what you said first. Let's say the "A" is a video, so we have row and column as the first two indexes, The red green or blue value as the third index, and frame number as the 4th index. So A(2,1,2,3) would be row 2, column1 of color channel #2 (blue) and you're looking at frame #3 (because the 4th index is 3). Now the blue value is one value -- it can't be a vector of 4 blue values. Please supply context as to what this 4-D array represents. When you send in indexes to A, what do those indexes represent?

[Bruno Silva]

Ok, I understand what you mean Image Analyst.

So for my specific application I am building a Neural Network to correct a curve fitting problem. The first dimension is the sampling rate, the second one the "window size" (how many values I check at once), the third one the number of hidden layers of that NN and the forth one the number of neurons of every hidden layer. What I would like to store (if I absolutely cannot store more than one value) is the Error for each combination of those (and MATLAB provides it through a variable).

I hope I was more clear now.

Answer 1

[Image Analyst]

You can store the error

In quadruple loop over all 4 parameters...

theActualValue = whatever(..........
thisError = SomeFunctionToComputeError(theActualValue, theExpectedTrueValue);
% Save error for this combination of parameters.
A(samplingRate, windowSize, numLayers, numNeurons) = thisError;

Answer 2

[James Tursa]

Here is one way to do what you have asked, although I don't really know if this is the best way to go about coding the real problem you are trying to solve:

D1 = 5:5:200;
D2 = [8 10 20 25 40 50 100 200];
D3 = 1:10;
D4 = 1:50;
A = @(a,b,c,d) [D1(a) D2(b) D3(c) D4(d)]

Then, e.g.

>> A(2,1,2,3)
ans =
    10     8     2     3

In this case, A is actually a function handle and not a 4D matrix. But it will yield the vector result you want when passed four inputs that are used to index into your "dimension" vectors.