Why is python so often used for machine learning if it's so slow?

why is python so often used for machine learning if it's so slow?

Attached: 1530742913792.png (506x547, 131K)

Libraries

Async + CUDA

you don't have to do all the intensive stuff in python, just control the logic. and yeah async and people who research machine learning might actually not be that into programming itself

Nonsense. Programming languages aren't "slow" or "fast".
It's just that the most popular Python implementation happens to run slowly.

wow. what a complete fucking retard.

*exhales loudly*
*Puts lips on bong*
*Inhales*
*Exhales cloud of smoke*
*Coughs loudly*
>DUDE LIKE..
>DUDE.
>PROGRAMMING LANGUAGES AREN'T SLOW OR FAST MANNNN

He's right though. Compilers / interpreters can be fast or slow, not languages.

Python isn't even a programming language ...

Most functions are implemented something, e.g. cpp or c. Python has very good data handling and a ton of libraries, making integration very easy.

If you actually implemented your ML algorithm in pure python it would, of course, run extremely slow.

That's very autistic of you to say.

All the hard work in Tensorflow and the like is handled in C++. Python just sets the parameters.

Languages can be designed in such a way as to elude the ability to be optimized.

And the one guy who is allowed to talk returns.

How isn't Python a programming language?
>inb4 scripting language
Retarded term only neckbeards use to arbitrarily define what is or isn't "real programming".

Attached: E825A88B-847D-4BF3-8C16-B881D5E57D86.jpg (1024x683, 53K)

Why was Common Lisp used so often for AI when it was slow on the contemporary machines of the 80s and 70s?
Because AI is hard, and usually accuracy and debuggability is preferred over raw performance. For time-critical applications, just get a bigger computer to throw at the problem and run your software as a service to clients.

Fuck, not Common Lisp, that didn't exist yet
I meant to say Lisp in general.
Lisp machines were an attempt to make lisp faster by making CPU's that natively process Lisp code.

Because the performance is acceptable. You don't need to learn how to program to be a data analyst.

Because Python is amazing for loading, arranging and processing data.
The real computing is done in C or C++ anyways.

The real computing is done in CPU byte code.

Not a *single* argument against.
I'm impressed. I knew Jow Forums was bad, but not this bad.

B-but Uriel told me C is close to teh hardware!!!1
Y-you mean CPUs d-don't run C natively? Are you crazy?