/sci/ - Science & Math

>>14751113
no
thermal engineer
no

your employer probably is not going to pay for MATLAB unless they have use for Simulink, almost everything else MATLAB does can be replicated with other languages that don't cost anything (Python + Spyder IDE is a very easy transition for MATLAB fags)

>>14751113
Just use python or julia, matlab is for first class faggots

>>14751113
> Do you use MATLAB out of academia?
no
> What is your profession?
A-OOOH! A-OoOH! A-OoOH!
> Are you just an at-home hobbyist?
no

https://www.youtube.com/watch?v=gI6sARmxEuc

>>14751113
I've used Octavia and Matlab at work for prototyping and PoC's.

>>14751117
he can just use octave, it is 95+% the same as matlab just free

>>14751117
MATLAB fag here, I have tried doing that python/s[pyder thing (took a course where the final project was to be made in python) and did not like it. python has benefits, no doubt, but it also has certain drawbacks and other idiosyncrasies which are very different from MATLAB. I know it is an industry favorite, but I would rather use octave or some other MATLAB clone if I had to

Matlab should rebrand as Simulink because the core program is a piece of shit. So to answer, no.

>>14751254
Could you list the most common things you do with Matlab? I imagine it's working with matrices, integration and maybe something else?
I tried using it once and found it to be terribly slow but maybe I was doing something wrong

>>14751254
which idiosyncrasies? i was a MATLAB fag that had to transition to Python and i didn't think much of it

i think the idiosyncrasies are more on MATLAB's end (indexing arrays from 1 instead of 0, output to console by default unless you use a semi-colon at the end of the line, can't remember more than that i haven't used it in 5+ years)

>>14751113
>>14751113
Yes, I use it almost daily in industry.
EE, DSP stuff
No, I use python at home since it's free.

I prefer Matlab. The code base is trustworthy (less update/compatibility issue compared to OSS), it has very good documentation, lost of integrations, performance is usually very good since many of the functions can be highly vectorized. The IDE is great for signal and exploratory analysis.

Well at least one Konica Minolta subcontractor uses it.
You need to know several technologies btw. You can learn most of them easily.

>>14751113
>and am trying to incorporate matlab into my corporate job
don't. that's a retarded idea. it costs lots of money, is poorly supported, and is a terrible programming language.
at worst use octave.
at best use numpy.
t. matlab expert that does shit tons of linear algebra

the only people i've encountered online that like matlab are aerospace engineers that don't understand control theory and rely entirely on simulink, and freshmen in college who have just encountered their first programming language.

You smooth brain cretin, python is infinitely easier,
and cheaper to implement in any meaningful way

>>14751274
samefag here, mostly I use it for processing of experiment data, fitting data and numerical modelling.
I cannot testify for its speed, for most of my tasks it is sufficient. like with many programming languages there are optimal and sub-optimal ways of doing the same thing, maybe you just chose the wrong way to go about it.

>>14751345
samefag here
>which idiosyncrasies?
matlab is more procedural, python is more object-oriented, so doing everything that is more than 1 step requires to get into a different way of thinking about how to do things. maybe it just takes some getting used to, but I did not like it.
namespaces are another thing I didnt like - in matlab everything is just there for the usage, in python its import, import import for every little thing which gets annoying. also sometimes it turns out that some functionality, even if it seems similar, is part of different packages for some reason. and even when you import a package, for some reason it does not import submodules which is very inconvenient if you need many submodules from a single package - scipy is very guilty of this.
oh, and the packages themselves. since matlab is made by a company, they have pretty good testing of their functions and toolboxes - I have yet to find some serious bug or broken functionality. python is open source and just about anyone can make a package, which sometimes includes shoddy work, uncaught errors, and broken functionality. I once even found a serious error in a well known package like numpy or scipy (cant remember which one was it).

other than that, the code structure is quite similar and the "idiosyncrasies on matlabs part" you mention are just a matter taking the time and getting used to the syntax, which I have not done sufficiently at that time

>>14751113
matlab is not the tool for a corporate workplace. Better off suggesting python, since it's a way bigger community, and some desk jockey will likely be more familiar with Python over some money sucking companies inferior programming language.

>>14751524
>it has very good documentation
i've never seen so many words say so little

>>14752743
Anon, F1 - it's so easy. Or CTRL+D. All the functions are well commented, most include examples too. That's not including the mathworks forums and online docs.

I have seen MATLAB everywhere I worked. You are jobless neet who is 30 years old that pretends to work. Matlab has Simulink ie the thing used to design anything control system related. From your car to the plane you ride. I can smell the jobless off of you.
>b-but free alternative
There is no free alternative to simulink.
Haven't personally used it since I am a test engineer but I have seen it everywhere I worked. If you want pictures to code for some real time shit its simulink or writing it yourself. I also seen it used for scripting to control emulators and little hardware testers.
Part of my job is getting sent MATLAB code from design that was used for sim shit and then converting it to C++ code which can work on ATEs that are used in production.

>>14752868
As good as Simulink is, Matlab itself is a piece of shit and is used less and less. Only boomers who settled on a tech stack in the 00s would rather use Matlab for things such as curve fitting instead of python
You work in an industry where Simulink is used, okay fine it's unparalleled there, but outside of that area if you take Simulink out of the equation you'd have to be retarded to use Matlab

I used Matlab to make some animated gifs of pdes for some conferences that I was presenting my dissertation work at. Even though I’ve been almost Python-only still since finishing my PhD, I still haven’t found a suitable library that replicates what to me seems like a pretty basic functionality. What I eventually started doing was drawing animations one frame at a time, saving the individual frames in a folder, and then passing them though ffmpeg or blender to turn them into a single animation. What a pain in the ass Python can be though.

>>14752932
Python is a programming language first, it wasn't designed as a scientific language.
But something like that can be scripted pretty easily
>Run your loop and export as many pics as you want
>Call ffmpeg from python
>Delete the frames automatically in python
I see that from the point of view of a scientist it can be a pain in the ass but that's hardly a weakness of python

>>14752932
Also it seems matplotlib has an animation library, what library were you using to make your pictures?

>>14752944
In matlab you can draw anything, tell matlab to save that frame, increment a time step of your animation, redraw, tell matlab to save the frame, etc. and then when you’re done tell matlab to save all the frames you’ve buffered into an animation. On the matplotlib side it only worked if you were making the kind of plot that matplotlib expected you to make, but if it was strictly outside the time evolution of one of the pdes they accepted then they didn’t have any straightforward way besides what I mentioned to do what I wanted to do

>>14751113
Yes
Electrical engineer, doing controls
No, at home I use whatever is convenient. Typically Python or C or AHK

Simulink is essential.

>>14752955
AHK? That’s kind of cool, what do you do with it?

>>14752907
Sure, you probably don't want to make a webserver in Matlab. However Matlab has a specific "niche". Even without simulink if you want to design digital filters its really nice. Has a lot of tools (that python doesn't have libraries for) that makes designing filters easy with visualization. I dont use matlab, since I am a test engineer, so i cant really give that many examples of use cases that makes sense but I'm sure they use it for all sorts of stuff that can't be done in Python as easily.

>>14752963
It seems matlab is great for engineering

>>14753006
matlab hobbles engineering

>>14753006
Mostly just for electrical engineers and computer engineers

>>14751117
MATLAB has better optimization libraries than Python. I learned it the hard way failing to solve a constrained nonlinear system of equations on Python for days and then succeeding on my first attempt on MATLAB. Sometimes you really do get what you pay for.

>>14751113
>am trying to incorporate matlab into my corporate job
why the FUCK are you doing that?

matlab is horrible legacy bloatware. use python and write proper code, you don't need to redo your entire pipeline end to end every 3 months like an academic.

>>14754907
and I say this as (currently) an academic. fuck matlab

>>14752955
>C or AHK
...when is that convenient?

>>14752963
>that python doesn't have libraries for
in the corpo world you're almost always going to be doing the same thing over and over again, so you can probably just write your own.

>>14754920
Yes, or you just pay a couple 1000 dollars for matlab licenses for your team and they dont have to maintain their own libraries and it just works. If you get one of your engineers who make 90k a year to work on libraries to replace MATLAB libraries and they spend 2 weeks on it well you just spent 3600 reinventing to wheel. Then you will have to maintain it. So more cost down the road.

>>14754939
matlab almost sunk a multimillion project i worked on because of its libraries that "just work"
fuck mathworks and their pileof subscription dogshit

>>14754939
>just pay for matlab and use matlab
but then you're stuck using fucking matlab.
>maintain their own libraries
unless your library is facing the web it doesn't really need much "maintenance" after you knock out most of the initial bugs. and usually what seems like a major task turns out to actually be fairly quick and easy to implement when you only have your own internal API to worry about.

you're not using matlab to build million-line enterprise bloat anyway, so most likely the typical talking points about reusability and maintainability don't really apply.

there is no perfect solution but for most rnd-ish things you can replace python with matlab if you're willing to give up slight conveniences like simulink. now if you're JUST using simulink or some other matlab-specific tool, it might make sense, but since OP talks about trying to goad the corpos into accepting shitlab he's most likely not an EE sketching out plant simulations all day

>>14755025
>replace python with matlab
*replace matlab with python

>>14751113
Anything that you can do in MATLAB, you can do in Python right now, more efficiently. This wasn't the case in early 2010s.

>>14751658
>freshmen in college who have just encountered their first programming language.
So the worst types of students anyway

>>14755047
you can't avoid having to understand control theory with pretty boxes in simulink, I guess. though I doubt that's what OP is doing.

more likely just a brainlet who has only written code in matlab and afraid to pick up another language.

>>14751113
I have never, personally, seen anyone use it in private business-oriented things. I have used it in both public and private research extensively, especially in any area where the military is related.
If you're going to do anything with MATLAB, you should learn Python, too. Not because Python is better - it's not, except at ML and AI, because a lot of those things originated in Python. You should learn Python, too, because Python is 95% as good at what MATLAB does but the majority of people use Python.
If you want to use MATLAB because you know where MATLAB is better than Python, I would recommend you take advantage of SMOP and similar add-ons to export your MATLAB code to Python so you can run them side-by-side and show your employer/RFI originator/grantor/etc. that MATLAB is the better option. Unless they are handcuffed to what is popular, you should be able to convince anyone who is serious about the project to let you do your own thing, since - if nothing else - you can show them that MATLAB is readily parsable to Python with add-ons to MATLAB. And that doesn't even cover that MATLAB's IDE contains a complete Python IDE in it that is accessible with keywords and vice versa.

If there's something that MATLAB or Python can't do, it's either because it's locked to some legacy system or it's outside the range of your expertise and needs an actual programmer/computer scientist to handle.

I focused almost entirely on MATLAB for 6 years. Not a lot of jobs seem to want it compared to shit like Python.

>>14751263
underrated post

I'll leave this here
https://profs.scienze.univr.it/~caliari/pdf/octave.pdf

Also this
https://www.youtube.com/playlist?list=PL1xkDS1G9As4evRC64aB788kx91CHdQ9W

>>14751113
>Do you use matlab outside of academia?
yes
>What is your profession?
electrical engineer, antennas
>Are you just a at-home hobbyist?
no

>>14755146
There are python alternatives like collimator.ai for that too now. But I think visual programming is a copout anyways. I work with nuclear scientists on the job the systems we're modeling are imperfect systems with past-input depending, variable time lag, that are more accurately modeled as autoregressive neural net models and we use PyTorch to do it.

>>14755025
>unless your library is facing the web it doesn't really need much "maintenance" after you knock out most of the initial bugs.
Oh, I've heard this one before

>>14755900
> I work with nuclear scientists on the job the systems we're modeling are imperfect systems with past-input depending, variable time lag, that are more accurately modeled as autoregressive neural net models and we use PyTorch to do it.
Please tell me you wont cause a nuclear disaster if the neural net goes haywire. You know like neural nets sometimes do.

Matlab is shit compared to Mathcad Prime

I do financial modeling. Despite having some pretty complex stuff we've developed, mostly clients want simple shit management can feel they understand so it's more simple visuals.

I used Matlab before for school and Stata. We use Python and R because we're a start-up and it's way cheaper.

We hardly use that anymore. A handful of people got interested in the complex models and white papers referencing Mandelbrot, etc. I literally mostly use Dax and M, Microsoft's piss simple data modeling and SQL language, because we mostly make simplistic BI reports or do ETL and let clients import data with a few basic merges on it into fucking Excel because that is what Boomer want.

We started hiring more design focused people and just teaching them this stuff because most people don't like complexity, they like Boomer friendly charts whose creation can be explained to them easily. lol.

>>14755950
If your library is 90% imports that have to be downloaded from public repositories then it's facing the web as far as I am concerned.

CPU architectures have insane backwards compatibility. If all your dependencies are static and you respect the API then your code does not need fucking maintenance aside from the long tail end of an exponential/lognormal bug distribution. You'll never get them all but unless you're making major changes frequently you will definitely get the vast majority of them eventually through sheer exposure.

This is /sci/, not /g/. Your 3.5kloc matrix wrangling library isn't the fucking linux kernel.

>>14755900
>simulink is copout
>we use autoregressive neural nets instead

>>14755950
you are actually dumb. scientific code does not need maintenance.

>>14756475
>CPU architectures have insane backwards compatibility. If all your dependencies are static and you respect the API then your code does not need fucking maintenance
>>14758081
>you are actually dumb. scientific code does not need maintenance.
Ahaha. Noobs out in force. I was like you, once. Starry eyed and naive. Just follow best practices and write good code, and it'll never break! It's Modular™ so you can reuse it! Oh my, to be young and hopeful again. I envy you. I really do.

>>14758632
>Ahaha. Noobs out in force. I was like you, once. Starry eyed and naive. Just follow best practices and write good code, and it'll never break! It's Modular™ so you can reuse it! Oh my, to be young and hopeful again. I envy you. I really do.
you literally do not know what you are talking about. you think the way it works in your discipline is how it works for everyone else, you are simply wrong.

>>14751129
>News today: Local /sci/ anon goes A-ooooh. A-oooh.
>public advised to stay at home.

>>14751113
The last company I worked at used matlab. Specifically the research team used matlab.

>>14752379
>in python its import, import import for every little thing
try a sagemath kernel

>>14751658
>octave

Lockheed, NG, and Boeing use it for all systems engineering. Python is a pile of garbage. Seethe motherfuckers!

>>14751113
I'm a civil engineer and I have used scilab a fair bit at my old job for fairly simple tasks.
We had some heat and humidity meters we set up all across buildings that had pretty bad readability, so I made a GUI that made it easy to compare different measurements, calculate dew points etc.
I have stopped working there and the code apparently stopped working after some update, maybe Octave is better?

>>14751113
yes
aerospace engineering
no

>>14751658
>aerospace engineers that don't understand control theory
lmao me

>>14751113
No, although I know people who do
Credit Risk Modelling & Stress Testing
No

For my job I mainly use SAS, with Stata only occasionally for panel models. The former due to the big datasets, the latter because the bank wants to have someone to blame in case of errors in the functions.
Still, I agree with >>14756351. The most used tool is still Excel and everyone at the top either wants a ppt or a spreadsheet which they can play around with. Indeed, SAS is just a tool to prepare a bit the data to be fed in Excel with a pivot table.

I did use Matlab extensively while I was a student, however if I did not care about the nice IDE and toolboxes I would currently use Julia.

>>14765164
any certs? do you use any other languages? I'm thinking of trying to learn fortran but I'm not that good with all the programming hoo-hah

>>14768504
No certs or anything. Only MATLAB usage at work. I got a gig doing simulation stuff out of college and have just been sticking in that lane for 6ish years.

A buddy of mine got a NASA job and it was mainly because she knew fourtran and cobol and few other weird old languages I had never heard of.

>>14768567
I really have to go down the xode route huh? I'm starting my masters in aero eng. next week and I only got 7 months of internship experience throughout my entire undergrad
FML

I work in the defense industry and many of the older PhDs love MATLAB. Some younger guys too. I assume this is because when you are at Uni in the US, the older PhDs want you to use what they are familiar with.

I work with someone who has a doctorate in Computer Science and they are in charge of developing an ML model. They did it all in MATLAB and then converted their model to Python. So fucking stupid IMO and it makes me cringe even thinking about it.

In my experience, MATLAB is good for some niche things like designing filters but Python can do most of everything else and is especially better for scripting that deals with files, reading sensor data, plotting, etc. I have been using Python for at least a decade now so I am probably considered biased.

>>14768590
I keep thinking about going back and getting my masters but just the idea curdles my soul

Simulink is used for control system/dsp automatic code generation

>>14751113
>>Do you use matlab outside of academia?
Are working free copies available?

>>14751113
The license is retardedly expensive for doing what other code and excel can do. Why do you want to?
>No
>Electrical Design
>I guess, not with Matlab though lol

>>14771291
octave tries to be free matlab. it's not a perfect implementation and is missing a lot of libraries. it's also a lot slower.
whenever i write code these days i make sure it will run on both though.
i absolutely cannot comprehend an academic, who's entire job is to develop and share ideas, would rely on something that is absurdly expensive to access. whenever i hear "but it's free because I'm a student" i just die a little bit.

>>14758632
>it'll never break!
that's not what I said, retard
>practices
doesn't matter. if nothing in the system changes it can be as big a ball of spaghetti as you want, the distribution of bugs you encounter will be lognormal or exponential with respect to time assuming you don't introduce new ones while fixing the old ones.
>modular, reuse
not what I'm talking about, moron. my system is full of programs and scripts I've written 5,6 years ago and haven't touched since - despite using them daily. because they don't touch anything that is constantly changing and I've worked out all the bugs.

>>14751113
Octave GNU
EE RnD
Also

Real men use gnuplot.

>>14771364
>system dies
>upgrade to macbook
>scripts say no