News for hackers. Stuff that matters.
Re:Show HN: A Genetic Algorithm I wrote in JavaScript to evo... by user24 on Monday December 13, 2010 @08:07PM (#2002678)
The code's fairly messy, but I hope the blog post and online demo will be enough to give the absolute beginner a good idea of how GAs work.

One day I hope to apply machine learning to the art of predicting real-life events. This is one small step towards that project, and I'd love to hear your feedback! - But it's 1AM now so I'll be offline for the next few hours. Here's hoping I wake up to a million upvotes! :)

Re:Show HN: A Genetic Algorithm I wrote in JavaScript to evo... by anathem on Monday December 13, 2010 @08:34PM (#2002751)
Related:

  * http://www.jurisgalang.com/2010/10/13/breeding-images/ 
  * http://www.jurisgalang.com/2009/02/01/breeding-strings/
Re:Show HN: A Genetic Algorithm I wrote in JavaScript to evo... by yters on Monday December 13, 2010 @10:56PM (#2003051)
Unfortunately there is something called the no free lunch theorem. Puts a significant constraint on using machine learning for real-life prediction.
Re:Show HN: A Genetic Algorithm I wrote in JavaScript to evo... by tel on Tuesday December 14, 2010 @03:14AM (#2003512)
Then again, all ML eventually is used for real-life prediction, so you ought to be able to buy a lunch if you really want one. The GA here is a fun start that'll introduce a training algorithm; there are plenty of gradient-aligned steps forward.
Re:Show HN: A Genetic Algorithm I wrote in JavaScript to evo... by yters on Friday December 17, 2010 @03:26PM (#2017628)
Sure, I'm just saying ML, just like its forebear AI, doesn't live up to the hype. Lunch is very expensive.
Re:Show HN: A Genetic Algorithm I wrote in JavaScript to evo... by tel on Saturday December 18, 2010 @05:09PM (#2020435)
Probably true. I'm not a big fan of the term ML due to the hype associated with it. Statistics however, which is pretty near the same thing most of the time, has proven itself quite often.
Re:Show HN: A Genetic Algorithm I wrote in JavaScript to evo... by Groxx on Monday December 13, 2010 @11:37PM (#2003139)
To that end, I propose a new acronym: The TISATATTANSTAAFLT (pronounced: "tiz-a tat tan staa fult")

The "There Is Such A Thing As The There Ain't No Such Thing As A Free Lunch" Theorem.

Re:Show HN: A Genetic Algorithm I wrote in JavaScript to evo... by user24 on Tuesday December 14, 2010 @05:15AM (#2003744)
You never know until you try ;)
Re:Show HN: A Genetic Algorithm I wrote in JavaScript to evo... by yters on Friday December 17, 2010 @03:26PM (#2017622)
Very true. I try jumping off of cliffs all the time cuz this time I might just fly!
Re:Show HN: A Genetic Algorithm I wrote in JavaScript to evo... by user24 on Saturday December 18, 2010 @12:52PM (#2019835)
Right. I'll just give up then. It's a good job no-one wasted time trying to apply machine learning to speech recognition, cuz there's no free lunch you know. Or handwriting recognition. Or housing prices. Or game strategies. Thanks for letting us know that there's no point searching for a solution via ML. Phew!
Re:Show HN: A Genetic Algorithm I wrote in JavaScript to evo... by IsaacL on Tuesday December 14, 2010 @09:08AM (#2004149)
Aren't genetic algorithms and machine learning distinct? I'm far from an expert, but my understanding is that with ML you're either trying to find a function (supervised) or clusters (unsupervised), whereas GA are a generic optimisation technique.

Maybe you could generate a bunch of different versions of an ML algo (say, ANN with different numbers of hidden nodes and starting weights), and use GA to evolve the best one. Doesn't seem like the best technique, though.

Re:Show HN: A Genetic Algorithm I wrote in JavaScript to evo... by user24 on Tuesday December 14, 2010 @09:25AM (#2004221)
Yeah, I've heard of using GA to find good weights for an ANN.

GAs are general optimisation techniques, but 'evolutionary computing' is very closely linked to AI and ML.