Anyone use Julia? How is it compared to MatLab?
Right now I just have it running in a terminal, but is there anyway to use it to graph equations?
Also how is it for calculus applications?
Anyone use Julia? How is it compared to MatLab?
Other urls found in this thread:
github.com
twitter.com
Just use python + matlab
Doesn't matlab need a license though?
My favorite language right now. It's way better than matlab and way better than python. I use it for mainly matrix stuff, graph/networks stuff, and other general algorithmic stuff. I don't really do symbolic arithmetic and I don't use julia for plotting but I'm sure there are libraries available for both.
But yeah it's a really nice language for actually doing things and I've coded in Haskell, O'caml, Matlab, R, Python, C/C++. Only thing I don't use Julia for is complicated data munging and plotting where R's ggplot2 and data.table is superior.
And Julia's linear algebra library is just ridiculously good. Way better than Matlab's, which is saying something.
I have a friend (EE phd) who thinks Julia is the best thing ever. He develops modules and stuff for it in his research. He believes that Julia will become the mainstream scientific computing language in a few years due to the JIT compilation and the growing support. The built-in package manager is also pretty nifty. Their Julia stuff was previously implemented in matlab.
You can use lots of python packages in Julia directly as I understand it. On the Julia site they give the following example for using matplotlib plt in Julia:
Pkg.add("PyPlot")
using PyPlot
x = linspace(0,2*pi,1000); y = sin(3*x + 4*cos(2*x))
plot(x, y, color="red", linewidth=2.0, linestyle="--")
Not sure what you mean by calculus applications. Linear algebra is obviously built-in with a pretty syntax and it has a myriad of scientific computing toolkits (not as many as python though as far as I know) including statistics, diff eq etc.
I've personally had issues getting recent versions of Julia running on my system due to all the dependencies, but that's what you get for running debian stable.
What does it have over python and R?
Yeah I'm CS phd and it really has the potential to replace the Matlab, Python, C++, Fortran mess we have in scientific computing right now.
I mean can it integrate, derive find limits etc?
My name is Julia :3 Actually it's spelt Юлия
There's a calculus package github.com
Indeed, that would interest me too. Also, are there libraries to read and write HDF/netCDF4 files?
Ph.D.? No shit.
I'm a senior undergrad sweng myself. But what is a postgrad doing here?
On the Internet you can be anything. But seriously are there any "somehow" trusty intellects here on /g?
Secrete question: how was 0! proven to be 1 (hint: using a Statistics theory).
>But seriously are there any "somehow" trusty intellects here on /g?
I've got a PhD as well, and I'm far from being an intellectual. All you need for a PhD is persistence. It helps if you're not dumb as shit, but being a genius isn't a requirement.
Interesting. Thanks for your take on this, persistence is a requirement indeed...
Out of curiosity, Ph.D. In what? Undergrad senior sweng here.
Autism has no age user
Lol yah if you got a phd you can't fuck around on /g anymore watch out guys.
Physics / Earth Sciences
Don't think a PhD doesn't take a toll, though. I've aged about 8 years while the PhD lasted only 3.5 years.
You'd be suprised at the amount of engineers and doctors who shitpost on Sup Forums
how is it better than matlab in your experience? Im especially interested in the linear algebra part. Is it simply faster?
Brutal. Brutal death metal.
I'm not sure what to say, but it's neat to have an intellectual neet on a Julia thread.
Lol jk about the neet part, however you or others may redicule getting a Ph.D, I'm sure yours was well earned. Maybe you should consider researching/teaching?
Also theoretically anything takes a toll, may it be responsibility, stress, or hardship of life...
get a proxy and tor, go to some russian site
you are welcome
Ok but I can just do pacman -S julia and be done with it.
It's funny but a lot of ee, cs, stats, math phds go and become code monkeys for google, ms, apple, fb etc. Only a small fraction of them continue doing research.
Wow, relentlessly depressing. So when you work for the big brothers you just trigger INC on their employee count?
I can think of one plus side of this, shit pays well. Maybe self-employment/contracting pays as good? idk.
I wouldn't say it's faster since Matlab has a lot of linalg functions that does multithreading implicitly, so it beats a lot of Julia functions that way. Julia uses BLAS too, so it also does do multithreading but not as much as Matlab. Hopefully, they'll implement it too since one of their goals is to come to parity with Matlab's linalg speed. For single threaded, Julia is usually the same speed or faster.
I mainly mean that matrices in julia are not complex floating point matrices like in Matlab. You can actually type the matrices as floats, ints, etc. Also extending to multiple dimensions is natural. Since it's typed, it picks up a lot of corner cases where Matlab would just return a complex vector or matrix. So a lot of things end up being consistent.
Other than linalg, the functions can be overloaded easily, which means we can extend svds, for example, to new types, given a couple of basic functions for the new types.
Plus it has macros and a neat interface for shell scripting. I could go on but no time.
Shit does pay well, especially if you're good at it. Contracting pays well too if you know the right people, friend of mine is an ee phd, he's getting 10K for a weeks work he's doing on the side. Pretty sure he could have asked for way more.
Pretty interesting, and encouraging. Chiefly for somebody who's stuck in a shithole called Syria.
Contracting with this local mass-retail IT guy (sort of like Walmart Jr.) to manage the whole facility. Getting $3k tops would be a great achievement... Even though we're building it from scratch with a relatively heavy security-mindset.
thanks. Im extremely interested in julia as a replacement for Matlab. Currently I'm solving large scale linear systems using various optimization techniques, so computational efficiency is important for me.
Do you usually have all the functions/toolboxes you need? I'm not sure I have time to start implementing toolboxes that already exist in matlab but not julia.
gnu octave might be better for your application...
lol
It has a nice interface for optimization, look up JuliaOpt and JuMP. But it's probably years away from having many of the toolboxes that Matlab has, if ever for some of the esoteric ones.
For eg, I know it has interfaces/APIs to computer vision, machine learning libraries and probabilistic programming libraries but it's years away from having it natively, not that having it natively is strictly required since the APIs just call C++ libraries but still.
Im currently working on a biotech library/toolbox for julia that has been implemented in matlab, python, R, Java and more.
This library is mostly based on complex lp/mip and bilevel optimizations of cells, and I was stunned that after optimizing routines a bit, my Julia code moped the floor of even R and nearly matched the performance of C code
For those doing repeated optimizations, do not use JuMP, the overhead of JuMP is appaling, but its syntax is praisworthy
What I think is lacking in Julia is a simple way to use parallelism, its insanely unintuitive and complex
Also plotting libraries are 'terrible' in the sense that no single plotting library has every feature I need, but together gadfly and pyplot are a good combo
What I absolutely love is the multiple dispatch, it makes a huge difference for my project to be able to take advantage of that feature
If anyone is interested in Julia ask ahead
I think Julia should have multi-threading in the new version, so multi-core parallelization should be way easier then.
I hope it does, but I think it has something to do w. the compiler
>but that's what you get for running debian stable.
Damn right.
Matlab student license is cheap. If you can't afford it you don't need it.
>his school doesn't provide matlab licenses
Of course they do for the labs. It's just if we want to buy for our own machines.
>It's just if we want to buy for our own machines.
Exactly.
Who is this semen demon?
mckayla maroney
license. hahahaha. paying for bits. hahahaha