/sci/ - Science & Math » Thread #11248139

36KiB, 648x361, 1_DcL9vLM9m-OPxXobLG-0Fg.jpg

View Same Google iqdb SauceNAO

Julia

Anonymous Mon 23 Dec 16:17:33 2019 No.11248139 View Reply Original Report

Quoted By: >>11248232 >>11248267 >>11248293 >>11248297 >>11248319 >>11248770 >>11249156 >>11249212 >>11249915 >>11250119 >>11252014

>muh amazing scientific computation language
>dataframes not included by default
>have to compile packages before use
>made by a company rather than full open source

Does it have any redeeming factors for scientific uses or is it just more vapourware?

Anonymous

Anonymous Mon 23 Dec 2019 16:49:36 No.11248206 Report

Quoted By: >>11248351

Bump. Heard about this in high school but never took the time to learn it. Any use in knowing it?

Anonymous

Anonymous Mon 23 Dec 2019 16:59:11 No.11248232 Report

Quoted By:

>>11248139
The image looks pretentious. That thing is only “advanced and scientific and cool” for people are not past their first year in science undergrad

Anonymous

Anonymous Mon 23 Dec 2019 17:13:04 No.11248267 Report

Quoted By: >>11250315 >>11250344 >>11251150

>>11248139
The real scientific languages are C and Haskell, all others are subpar in comparison when it comes to science. And there's also things like LaTeX, Scypy, Geogebra, Coq etc.

Anonymous

Anonymous Mon 23 Dec 2019 17:24:22 No.11248293 Report

Quoted By: >>11248311 >>11248541 >>11248890

>>11248139
It is as easy to write as Python and has built-in functions that essentially make it Numpy, SciPy, and MATLAB simultaneously while being as fast as C. Plus, it has all the tools you need for making Machine Learning models and neural networks with the Tensorflow Ecosystem. Models made in Julia can be saved and opened in Python so that other collaborators don't need to learn Julia if they'd rather stick to Python. It's also convenient for many things in and related to quantum science.

Julia has great potential for scientists in data-driven fields or fields that demand a lot of computational mathematics. Note that programming languages should be seen more like tools in a tool box and a scientist (or programmer) should have several of them for different purposes, with Julia being a contender for data-driven experiments.

Anonymous

Anonymous Mon 23 Dec 2019 17:25:54 No.11248297 Report

Quoted By: >>11249206 >>11250544

>>11248139
better than R, but then again R is a pile of garbage.
>takes eons to iterate over a loop
>syntax is trash

Anonymous

Anonymous Mon 23 Dec 2019 17:31:44 No.11248311 Report

Quoted By: >>11248342 >>11249297 >>11250045

>>11248293
Hum, no you dumb shit, Julia is definitely not as fast a C, what in the actual fuck. Also, numpy and scypy are far more useful, and don't even get me started with matlab. But the stupidest thing you said here was 'quantum science', like what the hell? Holy kek, listen, if you have no idea about a certain subject, restrain yourself from talking shit about it, ok? You've just embarrassed yourself hard.

Anonymous

View Same Google iqdb SauceNAO 1556572256072.jpg, 38KiB, 540x786

Anonymous Mon 23 Dec 2019 17:34:16 No.11248319 Report

Quoted By: >>11248344 >>11250045 >>11250347

>>11248139
>1-based indexing

Anonymous

Anonymous Mon 23 Dec 2019 17:40:21 No.11248342 Report

Quoted By: >>11249159 >>11249297

>>11248311
>Julia is definitely not as fast a C
https://julialang.org/benchmarks/

Anonymous

Anonymous Mon 23 Dec 2019 17:42:11 No.11248344 Report

Quoted By: >>11248864

>>11248319
>caring about indexing
ask me how i knoe you're a CS babby

Anonymous

Anonymous Mon 23 Dec 2019 17:46:01 No.11248351 Report

Quoted By:

>>11248206
It reached proper alpha only last year.

Anonymous

Anonymous Mon 23 Dec 2019 19:05:23 No.11248541 Report

Quoted By: >>11249228

>>11248293
>no Pandas

drrrrrropped

Anonymous

Anonymous Mon 23 Dec 2019 20:19:50 No.11248770 Report

Quoted By:

>>11248139
Julia is pretty awesome. It's almost as fast as c when you write optimised code but has syntax which is a combination of matlab and python. The only downside is the lack of libraries compared to python, although I'm pretty sure you're able to use many python libraries anyway.

Anonymous

Anonymous Mon 23 Dec 2019 20:58:10 No.11248864 Report

Quoted By: >>11249124 >>11249819

>>11248344
0 IS A NATURAL NUMBER

Anonymous

Anonymous Mon 23 Dec 2019 21:05:42 No.11248890 Report

Quoted By: >>11249109 >>11249297

>>11248293
>It is as easy to write as Pytho
>while being as fast as C
Is it actually true? Don't have to write optimized code to run it as fast as C? Is writing optimized code in julia as easy as writing regular code in python?
>Models made in Julia can be saved and opened in Python
How come? What do you mean by model made in julia?

Anonymous

Anonymous Mon 23 Dec 2019 23:08:59 No.11249109 Report

Quoted By: >>11249187

>>11248890
>Is it actually true? Don't have to write optimized code to run it as fast as C? Is writing optimized code in julia as easy as writing regular code in python?
Dude, c'mon, read what you wrote again and use your common sense, then ask yourself 'Is such a language possible?'

Anonymous

View Same Google iqdb SauceNAO Giuseppe_Peano.jpg, 54KiB, 580x734

Anonymous Mon 23 Dec 2019 23:16:07 No.11249124 Report

Quoted By:

>>11248864
NO IT FUCKING ISN'T

Anonymous

Anonymous Mon 23 Dec 2019 23:44:03 No.11249156 Report

Quoted By: >>11249159 >>11249305

>>11248139
Just use Gnu octave.

Anonymous

Anonymous Mon 23 Dec 2019 23:46:48 No.11249159 Report

Quoted By:

>>11249156
matrix statistics 10,000x slower
>>11248342

Anonymous

Anonymous Tue 24 Dec 2019 00:06:08 No.11249187 Report

Quoted By:

>>11249109
>ask yourself 'Is such a language possible?'
I don't think so, that's why I'm skeptical about the post I previously replied to.

Anonymous

Anonymous Tue 24 Dec 2019 00:18:24 No.11249206 Report

Quoted By:

>>11248297
>syntax is trash
Eh?

Anonymous

Anonymous Tue 24 Dec 2019 00:22:51 No.11249212 Report

Quoted By:

>>11248139
It's getting there. I've started using Julia for odd tasks that come up that aren't time-critical, so I can splurge away a day or two. Needs some getting used to, no question. Especially figuring out the use of array views etc. that give a massive performance drop if not used or not used correctly. Hard to say at this stage whether it'll become a great tool in the future, but definitely has the potential. Biggest market right now is applications where you'd normally write stuff in Fortran. Downside right now is that writing fast code in Julia isn't as intuitive as writing fast code in Fortran.

Anonymous

Anonymous Tue 24 Dec 2019 00:33:48 No.11249228 Report

Quoted By: >>11249309

>>11248541
Julia has a Python interface if you need to reach into Python and/or its packages. This also allows other collaborators to stick to Python on your code if they don't want to learn Julia.

Anonymous

View Same Google iqdb SauceNAO 1573527407973.jpg, 1MiB, 2836x2000

Anonymous Tue 24 Dec 2019 01:05:02 No.11249297 Report

Quoted By: >>11250876

>>11248890
Machine learning models from Keras, Tensorflow, whatever you're using:
>model = tf.Sequential(...)
>save (model, "Sequential Model")
That's all you need to do.

As for speed, check the benchmarks that another anon posted at >>11248342
>>11248311
>quantum science isn't a thing
Based retard, you have much to learn after high school graduation

Anonymous

Anonymous Tue 24 Dec 2019 01:11:12 No.11249305 Report

Quoted By:

>>11249156
Andrew Ng pls go

Anonymous

Anonymous Tue 24 Dec 2019 01:11:59 No.11249309 Report

Quoted By: >>11249328 >>11250372

>>11249228
Why not just use python then

Anonymous

Anonymous Tue 24 Dec 2019 01:24:21 No.11249328 Report

Quoted By:

>>11249309
Some operations aren't easily vectorized through numpy, and cython can be too much of a hassle. I agree that it's a crutch though, it's mostly intended for people who want to start with Julia but there isn't a Julia equivalent for some Python package.

Anonymous

Anonymous Tue 24 Dec 2019 06:17:09 No.11249819 Report

Quoted By:

>>11248864
No it wasn't. Including 0 as a natural number was something the burgers started doing in the 19th century and it was a big fucking mistake.

Anonymous

Anonymous Tue 24 Dec 2019 07:16:19 No.11249915 Report

Quoted By: >>11250989

>>11248139
I wrote quite a bit of Julia as part of a research project and generally I'd only recommend it to replace MATLAB and even then only because it's open source. It's a significantly less mature language than most which makes working with it painful. In particular there's no good development tools. Everyone in the community just uses Atom with some addons and says it's an IDE. The debugger is also highly dysfunctional which makes debugging code a fun time.

The speed promise is a fucking joke though. Compared to MATLAB it's a lot faster (not saying much since MATLAB is slow as shit) but it's nowhere near C. Generally code needs to be type stable to be really fast, so you end up writing code like it's a static language even though it's not and slapping ugly type annotations everywhere to ensure type stability. It can be fast, but it's a lot of work to make sure it's fast, probably more work than doing whatever you're trying to do in a static language.

Anonymous

Anonymous Tue 24 Dec 2019 09:50:07 No.11250045 Report

Quoted By: >>11250347 >>11251835

>>11248319
This. Developers don't know the difference between cardinal and ordinal numbers.

>>11248311
>matlab
Pretty sure that guy meant "matplotlib".

Anonymous

Anonymous Tue 24 Dec 2019 10:56:17 No.11250119 Report

Quoted By:

>>11248139
>Does it have any redeeming factors for scientific uses or is it just more vapourware?

it's an icebreaker to that one girl in undergrad that happens to be called julia

Anonymous

View Same Google iqdb SauceNAO 1391800135959.gif, 954KiB, 285x235

Anonymous Tue 24 Dec 2019 13:59:17 No.11250315 Report

Quoted By: >>11250428

>>11248267
>scientific
>Haskell

Anonymous

Anonymous Tue 24 Dec 2019 14:15:31 No.11250344 Report

Quoted By:

>>11248267
>scientific
>C
> all others are subpar in comparison when it comes to science. And there's also things like LaTeX, Scypy, Geogebra, Coq etc.
AHAHAHAHAHAHAHAHAHAHA, F99 > C.
>No native complex numbers
>multi-dimensional arrays? "just use pointers bro"
>implicit array reshaping? "bro just use integer arithmetic"

Anonymous

View Same Google iqdb SauceNAO 1561105896422.jpg, 503KiB, 989x750

Anonymous Tue 24 Dec 2019 14:17:28 No.11250347 Report

Quoted By:

>>11250045
>>11248319
clean, mathematical convention
>b-b-but my integer arithmetic
Cfags who never learned a real numerics language (Fortran).

Anonymous

Anonymous Tue 24 Dec 2019 14:31:57 No.11250372 Report

Quoted By: >>11250601

>>11249309
>Some operations aren't easily vectorized through numpy

Genuinely can't think of an situation where this would be the case

Vectors are vectors nigga

Anonymous

Anonymous Tue 24 Dec 2019 15:04:36 No.11250428 Report

Quoted By: >>11250449 >>11250514

>>11250315
Haskell is the most /sci/ language though.

Anonymous

Anonymous Tue 24 Dec 2019 15:19:01 No.11250449 Report

Quoted By:

>>11250428
>useless wank language that pseuds form a cult around because muh categories
sounds pretty /sci/ to me

Anonymous

Anonymous Tue 24 Dec 2019 15:51:02 No.11250514 Report

Quoted By:

>>11250428
dilate

Anonymous

Anonymous Tue 24 Dec 2019 16:06:05 No.11250544 Report

Quoted By:

>>11248297
>takes eons to iterate over a loop
Uh oh, he's retarded. You don't use for loops in R. Use combinations of vectorized operations.

Anonymous

Anonymous Tue 24 Dec 2019 16:29:57 No.11250601 Report

Quoted By:

>>11250372
Well, I can think of enough, and if you have to go down the route of running two nested loops in Python, you better hope it's a one-off calculation you'll never have to touch again.

Anonymous

Anonymous Tue 24 Dec 2019 19:20:17 No.11250876 Report

Quoted By:

>>11249297
>that picture
is this the highest possible level of mogging?

Anonymous

Anonymous Tue 24 Dec 2019 20:26:07 No.11250989 Report

Quoted By: >>11251049

>>11249915
>Generally code needs to be type stable to be really fast, so you end up writing code like it's a static language even though it's not and slapping ugly type annotations everywhere to ensure type stability. It can be fast, but it's a lot of work to make sure it's fast, probably more work than doing whatever you're trying to do in a static language.

So I'm still new to Julia, but isn't that pretty much the opposite what the "official guideline is". You don't need to put type annotations everywhere, but you need to make sure there aren't any hidden casts somewhere in your code that screw up performance.

Also, the #1 reason for me to even tinker with Julia and sticking to Fortran mostly is that pretty much no other languages have a from-the-start support for ND arrays with proper slicing.

Anonymous

Anonymous Tue 24 Dec 2019 20:53:13 No.11251049 Report

Quoted By:

>>11250989
It's true that you don't need to type annotate anything so long as the code is type stable, but in my experience it's extremely difficult to ensure type stability without type annotations everywhere, at least in code of any significant complexity. The type inference in julia isn't reliable enough to ensure that code is type stable without annotations.

The code base I was working on was ported from MATLAB and generally was horribly written so that might have something to do with the unreliable type inference, though.

Anonymous

Anonymous Tue 24 Dec 2019 21:39:00 No.11251150 Report

Quoted By: >>11251794

>>11248267
Poor post

Anonymous

Anonymous Wed 25 Dec 2019 02:22:02 No.11251794 Report

Quoted By:

>>11251150
Care to explain why (not that you can anyway), you fucking orangutan.

Anonymous

View Same Google iqdb SauceNAO df7.jpg, 90KiB, 680x907

Anonymous Wed 25 Dec 2019 02:48:32 No.11251835 Report

Quoted By:

>>11250045
I'm the original anon, and I did indeed mean MATLAB. Though, Julia has some decent visualization functions which would substitute for the matplotlib.pyplot stuff and then some.

Python protip: use seaborn alongside matplotlib if you're not already.

Anonymous

Anonymous Wed 25 Dec 2019 05:05:01 No.11252014 Report

Quoted By:

>>11248139
>Compshit
Not math or science.

Capcode	All Only User Posts Only Moderator Posts Only Admin Posts Only Developer Posts
Show Posts	All Only With Images Only Without Images
Deleted Posts	All Only Deleted Posts Only Non-Deleted Posts
Ghost Posts	All Only Ghost Posts Only Non-Ghost Posts
Post Type	All Only Sticky Threads Only Opening Posts Only Reply Posts
Results	All Grouped By Threads
Order	Latest Posts First Oldest Posts First

Your latest searches