Why is Python so fashionable in machine learning? - page 4

[Alexey Volchanskiy]

SanSanych Fomenko:

Here is a rubric by R

Here's a selection on time series

Here's a link to machine learning

Here 's R in Microsoft.

Here' s a selection in Russian

Here are the questions.

There are several packages on wavelets, like wavelets. When you open them up, there are links, and you can usually find books on how to apply wavelets to trading.

I had a whole collection, I can't find it at once, if I come across it, I'll post it.

R is mainstream, and if you can't find something, ask me, my knowledge of R is very limited, but obviously more than yours, I will help you.

My knowledge is approaching zero )) Ok, I need to find a package to calculate FIR filter coefficients. And not by window methods (long filters are obtained), but by Equiripple.

In Matlab, I just use such a utility, simple and convenient. I doubt there is anything like it in R. But still, please look for it, I'll mention you in the article ))

And you can write the code to calculate any filter with writing the coefficients to a file, that's what this utility generates. It only remains to add the part for writing the coefficients. It has been done for a long time and I won't describe it here.

function Hd = getFilter
%GETFILTER Returns a discrete-time filter object.

% MATLAB Code
% Generated by MATLAB(R) 9.3 and DSP System Toolbox 9.5.
% Generated on: 12-Apr-2018 22:48:35

Fpass = 0.2;   % Passband Frequency
Fstop = 0.3;   % Stopband Frequency
Apass = 0.01;  % Passband Ripple (dB)
Astop = 60;    % Stopband Attenuation (dB)

h = fdesign.lowpass('fp,fst,ap,ast', Fpass, Fstop, Apass, Astop);

Hd = design(h, 'equiripple', ...
    'MinOrder', 'any', ...
    'StopbandShape', 'flat');
% а из объекта Hd уже можно вытащить коэффициенты и много чего еще

***

[Alexey Volchanskiy]

Roffild:

Java (Scala) is the standard for distributed machine learning (Spark, MXNet, Hadoop).

R and Python have only linking modules to use these systems, not full-fledged support.

So I wondered why Google chose Python for its hardware instead of Java, for which so much has been developed within the Android project. Take for example the two IDEs, one of which is completely from scratch - Android Studio.

[forexman77]

By the way now went to the stackoferflow to ask a question. I spent an hour and could not send a message, then the paragraphs are not there, then something else.

Terribly stupid resource, for good reason I did not like it.

[Roffild]

TensorFlow is in C++, and only the API is in Python.

Spark can also execute Python code.

[Grigoriy Chaunin]

You can argue until you are blue in the face. Python is very common in the scientific community and is used for scientific calculations. It is, in fact, the standard for them. It is the standard for the ME. Macrosoft has customized their NS library for Python in the first place. Those are the facts. And why is this the case? Python is both a simple and powerful language with lots of libraries. That's why it has become popular for scientific calculations.

[Rashid Umarov]

I've never wondered about the branch question, but for me the answer is fully revealed in the hubs article - Python, Delphi and C++ through the eyes of a scientist / Habrahabr

It was also interesting to read the comments on the article.

[Alexey Volchanskiy]

Rashid Umarov:

I've never wondered about the branch question, but for me the answer is fully disclosed in the article on hubra - Python, Delphi and C++ through the eyes of a scientist / Habrahabr

It was also interesting to read comments on the article.

I read it, it's strange that such nonsense was published on hubrahabr. I guess the respected Rashid decided to make fun of it )) In general, it reminded me something of absurd arguments on this forum about harm of OOP).

A few stupid statements of the so-called "scientist":

You have no guarantee that the log file will be saved, due to the crash of the OS. (from me - it's his program that somehow crashes the OS!!!)
------------
The magic of the interpreter helps the good Python to see how much memory my number requires, and allocate exactly as much as it needs.

-----------
Windows code at least can get into memory area with OS code and crash it.
----------
But intuitive "naive" python code almost always works as expected. Similar code in pascal does not always work correctly. Well, naive C code almost 100% will not even compile without some special knowledge.
-----------

[Maxim Dmitrievsky]

They've had a lot of grief there, so they wrote an article :)

[Alexey Volchanskiy]

Maxim Dmitrievsky:
They've been there so much trouble that they even wrote an article :)

Foule, the martyrs of science)) they are given the wrong languages

[Rashid Umarov]

Alexey Volchanskiy:

I read it, it's strange that such nonsense published on the hubra. I understand that the respected Rashid decided to have a joke). In general, it reminded me something of absurd arguments on this forum about harm of OOP)

This is nonsense to some people, I was not joking. You just did not get it, as well as the "critics" of the scientist.