I stumbled upon that question.
I'm using python from matlab (reference). creating numpy arrays is as easy as py.numpy.array(rand(1, 10)). By passing the dimensions I can also create a small method that will allow me to pass multidimensional arrays.
The performance, though, is awful. I guess it's because the conversion is first to python's list and then to matlab array (boxing and unboxing).
Is there a way (like the one suggested in the referenced question) to improve that conversion?
