Slashdot Mirror
MATLAB Programming Contest Winner Announced
gooru writes "The MATLAB programming contest winner has been announced. It is a semi-annual programming contest organized by the MathWorks. What makes the contest truly interesting is the final phase is open source. Contestants may submit as many entries as they want and can tweak other entries."
The next contest should be to the death. Execute those who fail it. There can only be one. Then we'll see some real open source programming.
Comment removed based on user account deletion
Now, this Matlab contest is positioned to lead to the same silly cries. So, allow me to present a link to Professor Matloff's excellent article to head off any silly speculations about the decline of American technical prowess.
I do as much as possible of my work (bioinformatics) in Numerical Python. It's really nice to have the power of a general-purpose programming language as well as a numerical feature set that has equivalents for nearly every special-purpose MATLAB function I've ever needed. YMMV.
The correlation between ignorance of statistics and using "correlation is not causation" as an argument is close to 1.
Comment removed based on user account deletion
I guess for their speed programming award they are allowed to have prior source. If this wasn't the case, the author would have written it at 393 characters per second!
I'm beginning to wonder if this was rather some sort of PR effort rather than a true programming challenge.
If you're doing symbolic work, then Mathematica is the program to go with. But if you're doing numerical linear algebra and either don't need the speed of C/C++/Fortran or don't want to deal with those languages, it's kind of hard to beat Matlab. One nice combo is Maple/Matlab. Maple can call Matlab for numerical linear algebra work, and Matlab can call Maple for symbolic work.
Despite all of the people who complain about Matlab being unstable and using up resources, I've always found that running the command-line version of Matlab is fast and stable. The GUI version has some nice features, but they usually aren't essential to the work that I do.
On debian, apt-get search octav to see octave and extensions. Don't forget to install the additions octave-forge, etc. to get near-complete matlab equivalence. In some ways, it exceeds matlab, in some ways, it doesn't. And it is very compatible with matlab.
Octave has also been ported to MacOS X, and is available via Fink.
I agree, I have found octave *very* compatible; in my Quantum Mechanics class, we have frequent Matlab assignments, and I am able to cut/paste code directly between the systems, with no errors so far (but there was one Octave rendering bug with multiplot).
I don't know how Octave/Matlab stack up performance-wise for professional use, but for student use, it is ideal.
multifariam.net -- yet another nerd blog
I use Sci-Lab regularly. With Sci-Lab, I have no need to dole out bucks for the commercial version: Matlab.
All my MATLAB code is Open Source. And I am the most popular author (jointly with Luigi Rosa) this month. http://www.mathworks.nl/matlabcentral/reports/file exchange/top10Authors/
My Linux - (L)ove (I)s (N)ever (U)tterly eXPensive
Comment removed based on user account deletion
Comment removed based on user account deletion
Imagine a sandbox in which there are ants, sugar cubes, anthills, and rocks. Ants like sugar: collectively they want to bring as many sugar cubes as possible back to their anthills before sunset.
For this contest, you will write the control program that each ant carries with it. Ants, being so small, have some limitations, of course. Each ant can carry no more than one sugar cube at a time. Further, each ant can only see her local vicinity. Your program, which is run sequentially for each ant, knows only what that ant knows. Thus you must bring about the best possible global outcome based only on local conditions. The ants don't have any memory as such, but they can leave behind a chemical trail to guide themselves and others across the sandbox landscape.
Your score is determined by how much progress you make moving food towards and into the anthills. Ideally your ants will move all the sugar cubes onto anthills. Practically this may not be possible; do the best you can. You receive credit even by moving one sugar cube one step closer to an anthill.
Anyway, you want to find the shortest route that goes through n number of cities. I know in one variation of the problem you can't hit the same city twice, but I don't know if that constraint applied in this case. The ants leave a "pheremone trail" which evaporates after a certain amount of time. If the ants start out randomly choosing routes, but over time the routes with more software pheremone are reinforced, because the ant objects choose those paths preferentially.
Si la vida me da palo, yo la voy a soportar Si la vida me da palo, yo la voy a espabilar
Matlab is the Visual Basic of numerical computing -- a hodge-podge of grafted-on features. Yes, it gets a job done, yes it promotes code reuse because of the extensive numerical and graphing libraries, but as a "teaching language" it is weak on important concepts, and it is proprietary as all anything, turning engineering colleges into trade schools for MathWorks. And once engineering students glom on to it, you cannot, just cannot get them to use anything else.
I don't care if they implement a numerical algorithm in C++ or if they implement a numerical algorithm in Java -- both of those languages are pretty much callable from anything else on a wide variety of platforms. Yeah, you can call into Matlab too, but is there a free runtime you can download like with Java? And any kind of numerical algorithm using looping instead of built-in vector operations is going to be dog slow, so it is useless for any "production" use (in an academic environment, production use is where you throw a problem at it that taxes the capacity of whatever generation computers you have -- otherwise it is a toy numerical problem where everything you can discover with it has already been done.)
Comment removed based on user account deletion
When I last used Matlab, we used it just for the matrix calculator and, IIRC, it was free. When did it become a commercial product? Did I miss something or was just not paying attention back then?
"Lawyers are for sucks."
- Doug McKenzie
Care to explain how numpy is better than matlab? I'm not trolling, I'm actually about to start numerical processing for my research, and I'm stuck on the fence between using matlab (which we have on all our lab computers) and python, which i'd have the luxury of doing at home since it's free. i barely know either one, so i don't really have any loyalty to either side yet.
Basically, I like numpy because it's Python, and I like Python. More generally, I like having a general-purpose programming language when I'm writing real programs. MATLAB is good at crunching numbers and nothing else; numpy is almost as good (and in my experience, often faster) at crunching numbers, and it's also good at everything else Python is good at. String handling, DBI, and any data structure more complex than matrices, to name the examples I deal with most, are just unbelievably painful in MATLAB, IMO, but they're unbelievably easy in Python/numpy.
The correlation between ignorance of statistics and using "correlation is not causation" as an argument is close to 1.
Comment removed based on user account deletion
Comment removed based on user account deletion
A lot of these items are FUD.
1) You can choose any editor you want to write your matlab code. You just need to run it in octave. Since octave has a command line interpreter, you can show the result with any editor that can display the results of a run command (emacs will do this, too)
2) Yeah...it has readline, but that's about it.
3) Poster asked, besides being free...this is part of the price.
4) Not true. Any code not written in C, which is a good many of the numerical algorithms Matlab includes, have available source so that you can integrate the algorithms into any finished products (Matlab is for prototyping).
Other than that, you're asking for more than is really needed to extend the functionality.
5) Octave has a code repository. If they like what you write they use it. In other words, you can contribute to Octave.
6) Your fault/FUD. It took me about ten minutes.
7) I didn't have to. More FUD? Obviously this isn't a universal procedure.
8) I've never looked at my License file. I never track what it's doing. This has never been an issue.
9) See issue #3
10) Is this even a reason?
11) See issue #4
12) Obviously you don't have very good reasons. I will present some good reasons after we get through this.
13) This is true of Matlab as well. Try typing "ls" in Matlab and see what happens.
14) See issue #3
Having said all that, let me tell you why you should be using Octave.
The biggest reason is the free as beer thing. Matlab+ all packages needed is astronomically expensive. It's a big deal. We're not talking Microsoft-who-sells-to-consumers expensive - we're talking big-contractors-who-work-for-Engineering-firms expensive. It's kind of like the difference in price between Oracle and Postgres.
However, SOMETIMES it's worth it. As an Engineering student, I've tried and used regularly Matlab's image toolbox, Matlab's neural net toolbox, and their symbolic toolbox, and compared it to the normal canned algorithms.
Matlab is very, very good. They put an extra polish on every algorithm they write. In general, they're better written, and produce more clever results than anything else. Keep in mind that I was dealing with underconstrained problems, so the issues where matters of estimation. Matlab got more accuracy or faster convergence out of it's canned algorithms than you'd get if you wrote them straight from the descriptions supplied by the algorithm's authors.
Having said that, it's quite likely that there are certain areas that Octave will probably eventually fall behind. Symbolic work is one, I think, since their symbolic toolbox is actually an interface to Maple's symbolic engine, which they rent.
Maple doesn't have the manpower to compete with the OSS people writing computer algebra systems. IMHO, right now it's about tied. Three years ago Maple was ahead.
Mod me down and I will become more powerful than you can possibly imagine!
An open source matlab contest is the same animal as if Microsoft held an open source Excel or Visual Basic contest... except that Matlab costs a lot more, and Mathworks tend to be a lot more evil in its licensing terms.
Matlab costs about $3500... but at my work, somehow it costs $70,000 a year because of some weird ass licensing scheme matlworks sticks large government labs with. I've tired to convince my project that for that money it makes more sense just to hire programmers to add whatever features we need to octave and go tell mathworks to fuck themselves.
Oh, and by the way... all of that money is still not enough to get you bug reports noticed. For that you need to pay for some sort of premiere program.
There are 10 types of people in this world, those who can count in binary and those who can't.
um.. i=0:1:10.
Or do you mean address the first element with 0? Who cares if you start with zero, get used to it and move on.
If you use Matlab in a Unix enviornment ^c works.
You have dissed the language while it is amply clear that you have never used it. However let me clear this up - Matlab may not have the advanced features of c/c++, but it is designed to be a prototyping language - something that will help you test your algorithms fast. You can write code that will solve your differential equations, or do some signal processing with just a few lines - working with matrices becomes extremely simple since you don't have to worry about coding the intricacies of matrix manipulation. It makes FORTRAN look retarded as far as usability and speed of coding goes. It is definitely not as fast as programs created in C/FORTRAN, but it's not the speed of the code that's the objective here -it's how fast you can write up some code.
"When the only tool you own is a hammer, every problem begins to resemble a nail." - Abraham Maslow (1908-1970)
"""and yes, plotting is pretty bad -- py-gnuplot is the best available"""
...) by using your choice from a variety of backend renderers.
h tml
No at all. matplotlib is the current leader with excellent high-level functions in both an object oriented API and a MATLAB-alike functional API. Figures are rendered on screen with an interactive pan-zoom viewer or to a variety of a variety of file formats (ps, png,
http://matplotlib.sourceforge.net/
If you only need to render to file then the quality of PyX's output is unmatched. It includes very high level functions for common graphics operations and exposes much of LaTeX and PostScipt in case you need to do something detailed or unusual.
http://pyx.sourceforge.net/examples/graphs/index.
Anecdotal evidence: I work in remote sensing, mostly image processing apps using data from airborne sensors. The sensors are mostly experimental, and the image processing stuff is very researchy. We use MATLAB all the time- I honestly think if we used C++ (or pretty much anything other than MATLAB) during the algorithm development stages, it would take us literally 5-7 times as long to do anything, and in the end we would not have pursued nearly as many paths of research or done nearly as much analysis of our results. I don't know what it's like in other fields of engineering, but I know a ton of people who work in signal and image processing that would say the exact same thing.
I once struggled with one particularly tough problem for almost a year- I was developing in C++ at the time, since that was what the final implementation would be in. Again, this wasn't just implementing something that was well-understood...this was also researching new algorithms and doing lots of experimentation with lots of different datasets. After a year I got exasperated and and decided to develop the algorithms in MATLAB first. I reimplemented all the C++ stuff I had already done in the previous year, in about 1 month [!]. In another month I had a working system, much to my amazement. In another month I had tested and analyzed the living crap out of it in ways that I wouldn't have dreamt if I was still trying to do things in C++.
The only real competitor to MATLAB in this field is IDL (not to be confused with Interface Definition Language- you gotta love overloaded acronyms). In the open source world, you've got Octave, SciLab, and SciPy, but I think SciPy has the best shot at weaning people off MATLAB. So far I don't think any of the open source alternatives are quite mature enough for my purposes, but they're getting there. In my (rather specialized) field, Mathcad and Mathematica aren't contenders, for whatever reason.
In conclusion: as a programming language, it can seem a little odd (for instance, all variables are matrices). It's better to think of it as an interactive mathematical prototyping environment designed around a very specialized language. But if you have a need for that kind of tool, it can be a really amazing timesaver.
I think the grandparent post got it right. Matlab is designed for quick calculations, rather than final production code. If you're not sure that your method will work, it's much quicker to write a Matlab script to test things than it is to write a full blown C/C++/Fortran code. Additionally, I don't think that you're the target audience for the program. At best, the program is targeted to people who are applied mathematicians first, but it is often used by people who have no applied math background at all. Some of these people look at Fortran code, go glassy-eyed, give up, and delegate the problem (no matter how trivial) to a real computer science or applied math person. I've even heard (very intelligent) people complain about how hard it is to write Matlab code. Believe it or not, some scientists don't know how to write HTML, let alone Fortran; and sometimes these scientists need to get quick, reliable answers. Matlab is a great program for this type of application.
In most of my applied math classes (BA Computational & Applied Math, Rice U.) we had people from a variety of backgrounds who knew a variety of languages. Rather than force the Fortran people to learn C++ or force the C++ people to learn Fortran, we forced *everyone* to learn Matlab, which is a rather trivial task if you know anything at all about programming. But the main result of using Matlab was that programs that would take hundreds of lines could fit on a page or two. Instead of keeping track of pointers and variable types, etc. you could write concise, easy-to-read code. This saves a huge amount of time for the students doing the problem sets and for the graders. It also allows you to focus on the mathematical concepts rather than the mathematical or programming nitty-gritty.
Naturally, if you wanted to learn how to program in more depth, we had computer science courses available. And in the applied math classes where writing production-level, optimized code was more important, we usually used things other than Matlab. Although, often times Matlab would still be used for pseudocode or as a way of checking your answers.
I submitted the algorithm that MathWorks uses for generating the MatLab serial key.
It have a user friendly gui and everything, so I had hoped it had a chance.
Well guess what? I'm going to have to say you probably are wrong. I personally don't use Matlab but I support those that do in an educational setting. Our department makes extensive use of Matlab, we use it for instruciton, we use it for research. Now research groups espically are always money hungry. They always want more than they have so they try to save as they can. We have many who use Linux rather than Solaris to save on hardware and software fees.
None the less, we don't see Octave or Numerical Python used. Ever. In any group. It's Matlab or get out. The usage keeps increasing so the department will be scaling up the number of licenses we have. It's mostly core licenses as well. We do have some toolkits, but those are generally only a couple liceses for specific groups. Overall, it's the Matlab package that gets used.
Now I'm sure there are situations where these alternatives work, but I'd put money on the fact that they don't hold a candle to Matlab overall. Why? Well if they did, they'd be moving up, if not taking over.
I mean look at Firefox. HEre you are talking browsers, this is a market where you are dealing with lowest common denominator, users that have NO computer skills. Here the slightest change, even a purely cosmetic one can make someone refuse to switch. Yet, Firefox has been taking off in a big way. It probably won't become dominant for the reaons mentioned, but it's a serious threat.
Now one would think Matlab alternatives would be much better off. I mean here we are talking about techinal users. Electrical engineers use Matlab and so on. If you could offer something that's superior, free, and customizable, well the'd be jumping over in droves. Hell I'd expect it to be the standard, or at least on it's way.
Alas that's not what we see.
I personally see a lot of paralles to the audio world, something I work with personally. There's lots of OSS audio stuff out there, and there are those that push it as though it were all you need. However when I use it, I find it severly lacking. It cannot do what teh commercial software does. Even in many cases where it can, it does it far inferior either quality wise or ease of use wise or feature wise. It being free doesn't make up for what it lacks.
Octave uses GiNaC for its symbolic stuff. Currently, part of why Octave's symbolics aren't better, is not that the library doesn't do it, but that the "glue" hasn't been written.
What are you waiting for? :)
You've obviously never deployed nor administered a site license for MATLAB. Talk about a ROYAL PITA. Your time will come.
In the course of every project, it will become necessary to shoot the scientists and begin production.
Its the same argument that people give that still use Fortran. Most (scientific) departments have professors that make extensive use of Fortran for research and in thier instruction.
Its what the instructors learned originally and it 'can get the job done', but it is not neccesarily the best solution in whatever terms that you measure success. Professors use it because that is what they know. People don't jump all over new technologies because the original learning curve even if is 'superior, free and customizable'.
I think the same can be said of your department's use of Matlab. You have a set of users already using it with a set of code that already does a lot of things that they find useful; if they switched to another platform thousands of lines of Matlab code would need to be re-written. Since Matlad 'still works' they will never switch due to the legacy it has.
In my Machine Learning class at Portland State U, we've been using the Netlab toolbox from Aston University Neural Computing Research Group, which is a set of Matlab libraries and programs. I haven't used Matlab's own neural network tools or done any of this stuff in my working life, but NetLab is at least a good learning tool, and is itself GPL.
Several people in the class have speculated how much work it would require to port NetLab to Octave, but AFAIK nobody's actually taken a look. I downloaded it to my linux box but haven't tried to do anything yet myself.
From the Netlab page: "The Netlab library includes software implementations of a wide range of data analysis techniques, many of which are not yet available in standard neural network simulation packages. Netlab works with Matlab version 5.0 and higher but only needs core Matlab (i.e. no other toolboxes are required). It is not compatible with earlier versions of Matlab."
It's easier to be a result of the past, but more fun to be a cause of the future! http://www.spacefinancegroup.com/
That's an interesting package. I'll guess that it isn't as full-featured as the Matlab version, and in fact seems to have a somewhat different focus. I'd almost say it's a fairly decent complement to the Matlab toolkit, actually.
Naturally the big stumbling block for corporate development would be the GPL, but it is free, so one can't complain. Still, it'd be nice to see somebody compete with Matlab. A lot of times it's the toolkits, not base matlab, that ends up costing $$$$$. Usually when someone makes a good add-on toolkit, Matlab buys them out, as with Kohonen's SOM toolkit, but to his credit he does still release a GPL version.
There are some things in the Netlab toolkit that might do what I need for some current projects at work, though, so thanks!