Why is Python so fashionable in machine learning?

 

I read the articlehttps://habrahabr.ru/post/350042/, cool machine and again Google offers all API and development tools in Python. But why, it's slow, what's the point of cool hardware if you use a slow language?

Yes, I know libraries are written in pluses and they're fast. But the user code is still in python. I dabbled with python a long time ago, maybe something extraordinary happened over the years to make it so popular?

Who knows what, please write.

Бенчмарк нового тензорного процессора Google для глубинного обучения
Бенчмарк нового тензорного процессора Google для глубинного обучения
  • 2027.02.18
  • habrahabr.ru
Каждое устройство Cloud TPU состоит из четырёх «чипов TPUv2». В чипе 16 ГБ памяти и два ядра, каждое ядро с двумя юнитами для умножения матриц. Вместе два ядра выдают 45 TFLOPS, в общей сложности 180 TFLOPS и 64 ГБ памяти на один TPU Большинство из нас осуществляет глубинное обучение на Nvidia GPU. В настоящее время практически нет...
 
It seems to me that Python is just a language that "loosens up" people. Many things are "easier" with it. And the fact that it is "slow" is easier for us to get a more powerful machine than to bother with more complex languages, and even more so, with optimization...
 
Alexey Volchanskiy:

I read the articlehttps://habrahabr.ru/post/350042/, cool machine and again Google offers all the API and development tools in Pyton. But why, it's slow, what's the point of cool hardware if you use a slow language?

Yes, I know the libraries are written in pluses and they're fast. But the user code is still in python. I dabbled with python a long time ago, maybe something extraordinary happened over the years to make it so popular?

If you know anything, please write it.

In Python there are a lot of examples and forums where you can ask if you don't understand something. In R you have to figure it out yourself and it takes a lot of time and I haven't seen any forums on R at all (except one sub-forum).

Plus now there is the NumPY library. Vector calculations are much faster. But, also noticed all the same, that the code in the R console, in my opinion, is faster.

In general Python is much friendlier in learning and understanding, while R is richer and has more stuff in it in terms of machine learning.

 
forexman77:

In Python there are plenty of examples and forums where you can ask if you don't understand something. In R you have to understand everything yourself and it takes a lot of time and I haven't seen any forums on R at all (except one subforum).

Plus now there is the NumPY library. Vector calculations are much faster. But, also noticed all the same, that the code in the R console, in my opinion, is faster.

Basically nothing has changed Python is much friendlier to learn and understand, and R is richer and has a lot more in terms of machine learning.

R is not richer, all the machine trainers work in python, R is used by statistics and other uneducated people like local stone traders, because everything there is as easy as two fingers in three lines.

That's why there are so many libs and because every genius or student tends to do his or her own thing

 
Alexey Volchanskiy:

I read the articlehttps://habrahabr.ru/post/350042/, cool machine and again Google offers all the API and development tools in Pyton. But why, it's slow, what's the point of cool hardware if you use a slow language?

Yes, I know the libraries are written in pluses and they're fast. But the user code is still in python. I dabbled with python a long time ago, maybe something extraordinary happened over the years to make it so popular?

If anybody knows anything, please let me know.

you've asked this question 5000000000 times in different threads

just get used to it))

 
forexman77:

In Python there are plenty of examples and forums where you can ask if you don't understand something. In R you have to understand everything yourself and it takes a lot of time and I haven't seen any forums on R at all (except one subforum).

Plus now there is the NumPY library. Vector calculations are much faster. But, also noticed all the same, that the code in the R console, in my opinion, is faster.

Basically nothing has changed Python is much friendlier to learn and understand, and R is richer and has more stuff in it in terms of machine learning.

R is a slowpoke. It's true that I don't have much experience with it, I work with Matlab. And Matlab is also a retard)). They are all interpreters. As for friendliness, I've never noticed anything so special in my time.

I just don't understand why Google didn't use C++ or C# as a language for this device. Well, with Sharp it is somehow understandable, it's a language from MS, from the competitor. But what is the problem with the pluses?

 
Maxim Dmitrievsky:

you've been asked that question 5000000000 times in different threads

Just get used to it))

Like only one and never got any intelligible answer. All answers on emotion, like you have now. ))

I do not understand this. Google uses Java for Android. It would be logical to expect it in this hardware. But no.

 
Alexey Volchanskiy:

Who knows what, write plz.

Cython: C-Extensions for Python
  • cython.org
What users have to say about Cython: »You would expect a whole lot of organizations and people to fancy a language that's about as high-level as Python, yet almost as fast and down-to-the-metal as C. Add to that the ability to seamlessly integrate with both your existing C/++ codebase and your Python codebase, easily mix very high level...
 
Alexey Volchanskiy:

As for the friendliness - I did not notice anything so special in my time.

A lot of "googling" and it was more or less understandable in Python and more examples.

 
Alexey Volchanskiy:

I think there was only one and I didn't get any clear answer. All the answers are emotional, like yours now. ))

I already answered you 2 times in different threads

You seem to forget everything after a while

that python is a high-level language that is very convenient for working with vectors, matrices and neural networks

the speed is not important there because the main time consuming operations are done on the pluses and on video cards

all that's slow is preprocessing and is done only once

 
Python and R are the simplest languages to understand, in my opinion.
Reason: