If you all can help me with a poll. <https://twitt...
# datascience
t
d
From a quick look, I would guess NumPy and Pandas are fine, since they seem to be focused on more basic things. Particularly I would expect a matrix math library to be used, because think it's kind of a waste of time to have to explain "this is how matrix multiplication works, okay, now we're going to use this library because the important parts are written in C and it uses a non-naive algorithm so it's 10x+ faster than what you would casually write" It may be that parts of Pandas or NumPy do "too much", but you could just avoid using those parts.
👍 1
a
I don't think it is a good idea to reimplement matrix libraries and I would not recommend to teach people doing so. One one hand, basic linear algebra algorithms are well known and relatively simple. On the other hand, making them effective is really hard. Basically, you want a single super-optimized linear algebra library for each ecosystem and "from scratch" means making effective wrapper on top of it.
👍 1
t
Yes I agree on all these points from y'all. I think what I'm going to do is present both NumPy and non NumPy in the code files for class. But I'm going to teach with NumPy most likely. I was never going to reimplement a matrix library, but do loops manually for dot products and all that. There are people, especially programmers with no math background, that want to see things in pure code since they are not familiar with linear algebra. It's not optimal but it's a barrier that's removed for that particular crowd