CODEWITHSUNDEEP
multidimensional-arrayjulia
Norman
Norman

Reputation: 310

Julia: Generating Multidimensional Arrays by Broadcasting/Dot Operations on Vectors

I am trying to use broadcasting to store the value of function into a three-dimensional array. Below I use a simple example to illustrate what I am trying to do.

Let's say we have a function f that returns a value from three input values and three vectors X, Y and Z that store the input values:

f = (x, y, z)-> x^2+y^2+z^2
X, Y, Z = randn(100), randn(100), randn(100)

To evaluate f over all possible combinations of the values stored in the three vectors X, Y and Z, and then store the results in a three-dimensional array, we can do as follows:

[f(x,y,z) for x in X, y in Y, z in Z]

However, I want to avoid using for loops and use dot operators or broadcasting instead. So, I write the following:

broadcast(z->f.(X, Y', z), Z)

However, the problem with it is that the result becomes a one-dimensional array of two-dimensional arrays.

Is there an efficient way to evaluate f over all possible combinations of the values stored in vectors X, Y and Z and put the results in a three dimensional array without using for loops?

I don't want to use for loops because I am considering to put my code on GPUs in the future and it seems that GPU computing does not work well with for loops.

Thanks!

Upvotes: 0

Views: 307

Answers (1)

Bogumił Kamiński
Bogumił Kamiński

Reputation: 69819

The simplest is to add a third dimension to Z, e.g.:

f.(X,Y',reshape(Z, 1, 1, 100))

or directly generate data of appropriate dimensions:

X, Y, Z = randn(100), randn(1,100), randn(1,1,100)
f.(X,Y,Z)

Upvotes: 3

Related Questions