4 ms·
> Most things that are faster or more flexible than NumPy are much more alien and difficult to work with. Yes, this is a better statement of the optimal trade
by marmaduke 6y ago
> Most things that are faster or more flexible than NumPy are much more alien and difficult to work with.
Yes, this is a better statement of the optimal trade off in NumPy.
> same style of programming could be implemented in a library or language where those operations really do take advantage of parallel hardware
I think this is well vindicated by Theano, Tensorflow and APL before them.