I'll have to look into Octave, this is surprisingly the first I've heard of it. Based off of just the examples given in the article, it definitely looks like the better choice for ML applications. I think until Octave gets more community support, Python will continue to be the go-to however if for nothing more than the support it has.
Octave seems to be very generic. Why wasn't R considered? Isn't R a better option? It gives you good foundations for computational math and it also integrates with frameworks like TensorFlow. Also you get a fantastic tool like RStudio and you can even port matlab code with by using a matlab wrapper.
I would like to have included R as a candidate. I pretty much just ran out of time. If I can get to it one day I'll update the article.
python, and more specifically, keras on tensorflow and sklearn. R has great packages, as does matlab, and many other languages as well. But if you want to do any deep learning, tensorflow is the way to go.