We hate logic

Join a laid-back, close-knit community of mixed interests Get a free account!

halcy
99th percentile

No badges yet

Things.


halcy joined on Oct 29th, 2010, since that has made 21 posts that are still accessible today, 1 of which are threads. Helping shape the community, halcy has given 25 upvotes, and was last online on May 7th, 2013.


  • In What's the reason to teach Delphi before C?

    Well, mostly, two things:

    One thing is that often, teachers at high school (and the non-us equivalents of that) will only know one language or only know how to teach one language. Same goes for the people who actually design the curriculum. Changing these things takes forever.

    The other, probably more important thing is that C simply isn't a very good language for learning to program. There is much fighting over what constitutes a good "language for learning to program", but there is a fair bit of agreement that C isn't one. When programming in C, you have to be aware of all kinds of things that do not have very much to do with programming as a high level process - memory management, typing and casts, forward declaration, pointer arithmetic, the likes. These things are not too hard to learn, but when you cannot program at all, they are a distraction from what you are trying to do. Furthermore, in C, it takes very long to do relatively simple things, which makes for a fairly frustrating learning experience. A good "beginners" language, ideally, takes care of the details that are not related to program control flow and makes it easy to do simple but neat things. Python is such language; (people will probably kill me for saying that ) Perl is, too. Delphi... not sure, but you can probably do worse.

    What is important when starting to learn programming is not the details, not "How do you do specific thing x using language y": What is important is getting down the very basic principles of control flow, of what an "if-statement" and a "while loop" and a "for loop" really are - not how to write them down. What is also important is to actually program (Caveat: Dijkstra disagrees), and actually programming is only possible if you can, with some ease, make programs that do interesting things. You could not reasonably expect someone who is completley new to programming to do something like write a basic text-based adventure game after studying C for a week - Python, it's entirely possible.

    So, tl;dr: Because C is terrible for beginners and Delphi is probably not as bad. Also, bureaucracy.

  • In Artificial Intelligence

    @Fieyr

    Well, really, I doubt they solve any problem, there.

    I mean, I am by no means an expert for machine learning algorithms, but storage really isn't usually a problem these have. Processing power usually is. You have some classes, and you have an observation, wich you wish to classify. For that, you need to train a model. What that model looks like is going to depend on what you want to do exactly. Hiden Markov Models are a thing that is popular, Supprot Vector Machines are another. Sometimes, multiple methods are combined. The actual models do not "look" like the training data much. The training data, for offline training, is processed in one swoop, case by case, or needs to be iterated over all at once.

    Online learning, I do not know much about, except I vaguely remember Perceptrons were useful for that; but even for those, I doubt you'd need to keep the training data around much.

    Tangentially; digital computers are not brains and do not work like them, and it is probably not a good idea to try to make them work like them or emulate them. Especially when we do not know all too well how brains work, to begin with.

  • In Artificial Intelligence

    Realistically, most statistical models have fairly specific storage requirements that no thing not specifically tailored to them will satisfy very well.

    Sure, you could store training data in some nosql data store, but you could store it in a SQL based database just as well, or, heck, in a text file. You only need it to train the actual model, anyways, and that is a complex process, and usually done offline - for which you need all of the training data, usually at once, not much querying required there. It's hard to train anything but the simplest models dynamically - and perfect training is often NP-hard, anyways.

  • Zero Gravity Cats

    Cats, in zero gravity. That is all.

  • In Poll: What is your age group

    Saa... 23, 24 in less than a week. I feel old.

  • In Starcraft 2

    Despite being from the EU, I play on the NA server, due to friends there, I guess.

    halcy 500, but people seem to have trouble finding me, so mail: lorenzd+bnet@gmail.com

    (I suck, but if you feel like playing, lets~)

  • In She is alive!

    hooray, markov-model based text generation

  • In Post what you're listening to.
    www.youtube.com/watch?v=ROwH8ZEST1U

    "Catgroove", Parov Stelar. Electroswing. Excitingly danceable.

  • In Show off your desktops!

    selfmade SVatG background, lol

  • In What do guys want to see in girls?

    If I had any idea about that, I'd sure have one problem less.

    See, people think they know what they want, and then they fall madly in love, and then they get bored after 3 months and don't know what just happened, or they're just not sure about this or that person... it just takes time, I guess, time and trial and error.