Since the Explorer team doesn’t seem to be interested in adding random variable sampling or advanced scientific functions, I have started my own project inspired by SciPy with the goal of adding efficient computation over multidimensional arrays in Elixir. The code is based on bindings to the Rust library ndarray and supporting packages. I’m linking it here in case anyone finds it interesting.
Forum topic here: SciEx - scientific programming for Elixir (based on bindings to rust's ndarray)
The following snippet shows an example of what’s possible (keep in mind this in the very early stages):
iex(4)> a = SciEx.Random.draw_from_normal(0.0, 1.0, 10)
#SciEx.F64.Array1<[0.980546308591242, -0.4124865578379766, -0.2678229458353892, 0.13895119159920113, -0.05895938590440134, 0.5463383504051798, -0.47139019152800626, 1.1432875805108502, -1.053279376911445, -0.4890897953025381], shape=[10], strides=[1], layout=CFcf (0xf), const ndim=1>
iex(5)> b = SciEx.Random.draw_from_normal(0.0, 1.0, 10)
#SciEx.F64.Array1<[-0.6112979257481231, 0.04187616378275234, 0.024658043411414837, -0.7423086166282216, 2.5287898587121993, -1.5968438074887625, 0.3028680311098233, 1.1657682958696634, -1.3245816353940532, 1.1693983808728805], shape=[10], strides=[1], layout=CFcf (0xf), const ndim=1>
iex(6)> use SciEx.Operators
SciEx
iex(7)> 0.5 * a + 0.7 * b
#SciEx.F64.Array1<[0.06236460627193485, -0.17692996427106167, -0.1166508425297042, -0.45014043584015456, 1.7406732081463385, -0.8446214900395438, -0.02368747398712684, 1.3876815973641894, -1.4538468332315597, 0.5740339689597472], shape=[10], strides=[1], layout=CFcf (0xf), const ndim=1>
