What would be the appropiate language for scripting combinatorial algorithms? Performance is important because the complexity of my problems always is exponential and higher.I tried Python but it was too slow and used too much memory unless the algorithm was already mostly implemented in a specialized library. C is good, but the writing is quite slow, and I have to reinvent the wheel for most applications. Examples of things that I need to do:1. Generate all the partitions of {1,2,...,n} satisfying some property.2. Generate all the graphs that satisfy a given condition.3. Concatenate a small collection of integer-valued lists a large number of times, in mostly random order.So, the typical thing would be to iterate through a huge set of combinatorial objects and perform some memory moves and comparisons on their data.
Wolfram Language
>>101554081This is exactly what I'm doing, lol. Solving the TSP right now. Anyway, the answer is either C or C++. You can wrap them from Python no problem; it's especially easy in C. C++ has a lot of stuff built in, but if you want the absolute highest performance, that may be less helpful that you think. A lot of the things in its standard library are actually pretty slow, whether it's due to backwards compatibility or upholding properties that you don't care about in your algorithm. So you may still end up writing custom stuff or bringing in external libraries. For the highest performance, you'll probably want lots of control over memory allocation and deallocation too, so you won't use all the "modern" patterns. For this kind of code, the most helpful part of C++ is templates, really.
>>101554762This
>>101554081I would also like to get some other opinions on that. If you want to do symbolic calculations, Maple. Literally every other math focused language sucks, in particular Mathematica aka wolfram and sagemath. I've heard good things about the symbolic julia package, but haven't tried it enough to say too much about it.
>>101554081Java
>>101554081common lisp
>>101554081I've used c++ for a lot of dynamic programming and optimization work and it was probably the most straightforward. Yes Python is easier to write than c++ you are right but all the functions are obfuscated and the libraries honestly suck.
prolog
>>101554081still C
>>101554081C. C++ if you need templates.With CUDA if you need parallelism.
>>101554081this is an interesting thread, why does g let it reach page 10 but e celeb drama last forever
>>101559746Most /g/ users are retarded. The board would be improved if 80% were banned.
>>101559746/g/ is for people who got filtered from (the on-topic part of) /sci/
>>101554081linear algebra
You are looking for linear programming. Its not as gay as it sounds, its actually pretty cool once you get the hang of it.
fastlisp is combinator based
>>101560205Yeah, linear programming can come in really useful for combinatorial optimization problems, but it can be kind of a pain in the ass to implement.
>>101555872there is literally nothing good that's oss. fuck stephen wolfram for making his shit closed, dumbass british schizo
>>101560490What is fastlisp? Fastest Lisp I know is SBCL, and that's just a normal dynamically typed lang.
>>101560054>(the on-topic part of) /sci/tell me more about this, I only see sci being used for IQ and vaccine threads, I never see actual algo theory discussed
>>101560851well I'm not in /sci/ am I?ask there
>>101559746/g/ is for people who got filtered from /ohm/.
>>101560851>>>/sci/16294582
>>101560911I thought ohm was from diy tho
>>101554081C++
>>101554081>Performance is important because the complexity of my problems always is exponential and higher.C or C++.
>>101554081Haskell!
>>101559746How the fuck is it interesting? OP asks retarded question, replies to the thread say nothing of substance. All the (You)'s this post received are literal fart sniffing copium on top.
What exactly are you trying to do? If you’re devising new algorithms for your own use from scratch and is comfortable with python, take a look at pypy, it helps a lot on loop-heavy code.If you’re building a package of new algos to be used by others, you probably want to use c++ as it’s the lingua franca for high performance, foundational softwareBut if you’re implementing existing algorithms to your problems, it’s extremely likely that there already packages implementing them, most likely in c++, but occasionally you’ll see new cool stuff in python, Java, R and other academically popular languagesMost likely your problem can be translated to integer linear programming, though, and if it’s solvable on modern hardware it’s worth the effort to learn how to use specialized software. GLPK and HiGS are popular open source packages for that, highs even being a dependency on scipy for a few years nowIf you’re just doing homework and want something more fun than strictly performant, SML can be learned in a weekend and as all ML family languages can be fun to solve combinatorial problema with recursion and pattern matching. Also quite fast if you compile the code. Same for other languages of the family, like ocaml, f#, haskell and more recently flix
>>101564835Flix? Never heard of that.
>>101564835P.S. >>101565370Can you redpill me on that? Why would I use it over other ML-family languages?
>>101564835I research in combinatorics, so I need to experiment with really specific settings. I gave some examples in the first post, but if you want something more concrete, the last time I had to generate all hypergraphs of certain type and checked if a formula relating the number of hyperedges and vertices was true. So, it is not combinatorial optimization.
>>101566487Any particular reason not to use the usual languages for high performance algos, C and C++? Is it the need to do manual memory management? If so, try Java. Once it warms up, it's one of the fastest GCed languages for algorithms. Just be mindful of the boxed/unboxed distinction.
>>101554081>>101566487>>101566595Depends on whether you can recast your graph algorithm as a matrix algorithm (e.g. BFS is also a matrix algorithm), in which case you could even use pic related and call LAPACK, or parallelise the shit out of it if possible.Otherwise I see no reason not to do C++.
>>101554081>C is good, but the writing is quite slow, and I have to reinvent the wheel for most applications.C++ has everything you need.
>>101566595>>101566743>>101566761So C++ is the way.Thank you guys
