Channel 9 Forums - Coffeehouse - Machine Learning http://mschnlnine.vo.llnwd.net/d1/Dev/App_Themes/C9/images/feedimage.png Channel 9 Forums - Coffeehouse - Machine Learning http://channel9.msdn.com/Forums Channel 9 keeps you up to date with the latest news and behind the scenes info from Microsoft that developers love to keep up with. From LINQ to SilverLight – Watch videos and hear about all the cool technologies coming and the people behind them. http://channel9.msdn.com/Forums en Thu, 23 May 2013 22:31:08 GMT Thu, 23 May 2013 22:31:08 GMT Rev9 17 -17 -1 Coffeehouse - Machine Learning Anders Heijslberg answered a question in this video about what he thought was the "next big thing" in programming language and software and he said he thought "machine learning" was the most pressing issue.

His answer was the same as a few years ago when concurrency was "the elephant in the room". Have you any experience in data mining heuristics and do you agree that this is the next "nut to crack?"

How would machine learning be implemented in a language like C# or Visual Basic, and what application to you this this may have.

Mike Swanson had a video on Genetic Algorithms a while back, and Josh Smith has implemented this in an objective c example.

Is this juts a case of a few people needing to solve a specialised problem or could there be more? I dealt with an American marketing company a while back that was charging many thousands of pounds in order to generate "the right customers" and if I had the time, I would like to have a go at implementing this paper from MIT as a lot of money is beginning to rely on such expertise.

What remains is for someone to explain how the .NET Framework or C#/VB could have this as a first class language or library construct?

]]> http://channel9.msdn.com/Forums/Coffeehouse/Machine-Learning/0df8303499344f159aeda03800f44e48#0df8303499344f159aeda03800f44e48 Thu, 19 Apr 2012 14:49:29 GMT http://channel9.msdn.com/Forums/Coffeehouse/Machine-Learning/0df8303499344f159aeda03800f44e48#0df8303499344f159aeda03800f44e48 Vesuvius 17 http://channel9.msdn.com/Niners/vesuvius/Discussions/RSS Coffeehouse - Machine Learning Machine learning will be used to mine all the data out there that is being generated at an unbelievable rate, detect patterns and extract signal from the noise ... so it's a huge thing for the future.  Anders was saying that programming language designers don't know what to do in the languages itself to support machine learning, yet.

However, in the same Lang.NEXT conference, there was a presentation - Reverend Bayes, meet Countess Lovelace: Probabilistic Programming for Machine Learning -  on machine learning and the presenter addresses Anders' statement (as I told exoteric in another thread). 

So, that brings us to libraries, like Infer.NET and F# using it, makes a great start.

 

]]> http://channel9.msdn.com/Forums/Coffeehouse/Machine-Learning/87a8cef9f7e24d07bd08a03800f7d2b8#87a8cef9f7e24d07bd08a03800f7d2b8 Thu, 19 Apr 2012 15:02:17 GMT http://channel9.msdn.com/Forums/Coffeehouse/Machine-Learning/87a8cef9f7e24d07bd08a03800f7d2b8#87a8cef9f7e24d07bd08a03800f7d2b8 Richard Anthony Hein 17 http://channel9.msdn.com/Niners/Richard.Hein/Discussions/RSS Coffeehouse - Machine Learning It seems like machine learning really begins to shine when you have a specialized use case. So, implementing it in a general purpose framework or language doesn't make sense on the surface.

However, if you could figure out some of optimizations that have been made to make .NET, the CLR, and the languages work for general purpose use, and adapt them so that they can learn more about what you are trying to do, and adapt to make execution more efficient for the task at hand. Obviously that was a very vague idea, and it becomes very tricky to have code automatically adapt when it's already been fine tuned to work well in certain cases like arithmatic or string manipulation.

]]> http://channel9.msdn.com/Forums/Coffeehouse/Machine-Learning/9a15a4183602493b858aa03800f988f6#9a15a4183602493b858aa03800f988f6 Thu, 19 Apr 2012 15:08:31 GMT http://channel9.msdn.com/Forums/Coffeehouse/Machine-Learning/9a15a4183602493b858aa03800f988f6#9a15a4183602493b858aa03800f988f6 kettch 17 http://channel9.msdn.com/Niners/kettch/Discussions/RSS Coffeehouse - Machine Learning Yeah, machine learning seems to apply more to the libraries and maybe compilers, then language elements.  However, putting support for things like random or sample (a proposed keyword for Infer.NET) might make sense.  The way F# integrates with Infer.NET in the presentation shows that there's some room for language support.

 

]]> http://channel9.msdn.com/Forums/Coffeehouse/Machine-Learning/681799e51957436d93f7a0380109f329#681799e51957436d93f7a0380109f329 Thu, 19 Apr 2012 16:08:17 GMT http://channel9.msdn.com/Forums/Coffeehouse/Machine-Learning/681799e51957436d93f7a0380109f329#681799e51957436d93f7a0380109f329 Richard Anthony Hein 17 http://channel9.msdn.com/Niners/Richard.Hein/Discussions/RSS Coffeehouse - Machine Learning @Richard.Hein: Downloaded  the video in the link, and the one at infer.NET so that's my evenings viewing sorted.

]]> http://channel9.msdn.com/Forums/Coffeehouse/Machine-Learning/f48214312af74759add6a038011f3b51#f48214312af74759add6a038011f3b51 Thu, 19 Apr 2012 17:25:46 GMT http://channel9.msdn.com/Forums/Coffeehouse/Machine-Learning/f48214312af74759add6a038011f3b51#f48214312af74759add6a038011f3b51 Vesuvius 17 http://channel9.msdn.com/Niners/vesuvius/Discussions/RSS Coffeehouse - Machine Learning I like this. Machine learning should be the focus of the software industry. Enough coming up with new and pointless ways to put graphics on a screen, or put data in a data store.

]]> http://channel9.msdn.com/Forums/Coffeehouse/Machine-Learning/527271b3d25142b9aeb5a039016c970b#527271b3d25142b9aeb5a039016c970b Fri, 20 Apr 2012 22:07:25 GMT http://channel9.msdn.com/Forums/Coffeehouse/Machine-Learning/527271b3d25142b9aeb5a039016c970b#527271b3d25142b9aeb5a039016c970b Bass 17 http://channel9.msdn.com/Niners/Bass/Discussions/RSS Coffeehouse - Machine Learning Machine learning is a new focus for language designers (or will be, as Anders stated in the panel referenced above). How/when is still unclear. In some sense, from a tooling perspective, one application of machine learning would be to automatically generate code (like a function) based on some specified (declarative) input from users that provides enough information for reliable inference (so, here's what the function can do, now make me a function).

We'll keep a keen eye on this (in addition to more related conversations on C9 with folks doing this type of work like Andy Gordon and Erik Meijer...).

C

]]> http://channel9.msdn.com/Forums/Coffeehouse/Machine-Learning/6c4f236ce35d483c8edfa03a014a00d8#6c4f236ce35d483c8edfa03a014a00d8 Sat, 21 Apr 2012 20:01:30 GMT http://channel9.msdn.com/Forums/Coffeehouse/Machine-Learning/6c4f236ce35d483c8edfa03a014a00d8#6c4f236ce35d483c8edfa03a014a00d8 Charles 17 http://channel9.msdn.com/Niners/Charles/Discussions/RSS Coffeehouse - Machine Learning I've been studying this topic lately too. I'm currently taking online courses in Machine Learning (Andrew Ng) and Probabilistic Graphical Models (Daphne Koller) from Coursera and would recommend these as good introductions to the topic. There's a Vision class starting soon too!

 

 

]]> http://channel9.msdn.com/Forums/Coffeehouse/Machine-Learning/5e73678fadba4be29e3fa03a015ce72e#5e73678fadba4be29e3fa03a015ce72e Sat, 21 Apr 2012 21:10:19 GMT http://channel9.msdn.com/Forums/Coffeehouse/Machine-Learning/5e73678fadba4be29e3fa03a015ce72e#5e73678fadba4be29e3fa03a015ce72e contextfree` 17 http://channel9.msdn.com/Niners/contextfree`/Discussions/RSS Coffeehouse - Machine Learning I'm not 100% sure this use case applies, but since we are talking about mining data, I guess in some sense it applies...

I remember when Quake was the state of the art FPS and they allowed 3rd party modules (scripts, don't remember the details) to be added to the game engine. There were many bot scripts that allowed you to control the bots using these add-ons. Some were getting to be so good that sometimes you could not tell whether it was a bot or a real person playing.

Sadly the bots you find in today's games are extremely stupid. For instance, in Gears of War (1/2/3) the bot behavior is so predictable that you can quickly learn how to defeat them. It seems game developers gave up on making truly smart bots.

So I was thinking whether it would be possible to have learning bots. Basically, you start out with a "blank" bot that just stands there. When he gets shot, he learns the movements from the opponent that killed him. So this is added to his database of moves. As he gets into more combat, he becomes smarter and smarter.

I know that translating the moves of the opponent to work in a generic way in a dynamic environment is probably quite complex, but I believe much more difficult problems have been solved in the past already.

]]> http://channel9.msdn.com/Forums/Coffeehouse/Machine-Learning/343a8a0cc6ae4149bab1a03a017ec007#343a8a0cc6ae4149bab1a03a017ec007 Sat, 21 Apr 2012 23:13:33 GMT http://channel9.msdn.com/Forums/Coffeehouse/Machine-Learning/343a8a0cc6ae4149bab1a03a017ec007#343a8a0cc6ae4149bab1a03a017ec007 BitFlipper 17 http://channel9.msdn.com/Niners/BitFlipper/Discussions/RSS Coffeehouse - Machine Learning

, contextfree` wrote

I've been studying this topic lately too. I'm currently taking online courses in Machine Learning (Andrew Ng) and Probabilistic Graphical Models (Daphne Koller) from Coursera and would recommend these as good introductions to the topic. There's a Vision class starting soon too!

 

Thanks for the heads up, I had looked at my local University, and they only had a course using machine learning with Games. I will definitely do the course. How long has the Andrew Ng been running? Will i have to wait a few months before the next course becomes available?

]]> http://channel9.msdn.com/Forums/Coffeehouse/Machine-Learning/e05c0696ea244bf7842da03b00aa0460#e05c0696ea244bf7842da03b00aa0460 Sun, 22 Apr 2012 10:19:00 GMT http://channel9.msdn.com/Forums/Coffeehouse/Machine-Learning/e05c0696ea244bf7842da03b00aa0460#e05c0696ea244bf7842da03b00aa0460 Vesuvius 17 http://channel9.msdn.com/Niners/vesuvius/Discussions/RSS Coffeehouse - Machine Learning That page is kind of confusing - where it has

CategoriesDurationNext Session
Computer Science
Mathematics and Statistics		10 weeks23 April 2012

 

, the "Next Session" is actually the beginning of the entire class (it's referring to the whole course duration as a "session"), i.e. it's when the "10 weeks" starts. (The previous session, which is what I went through, was last fall.) So it actually begins, erm, tomorrow - should probably sign up now ...
]]> http://channel9.msdn.com/Forums/Coffeehouse/Machine-Learning/73c0740925e54f10b938a03b00bd1d6c#73c0740925e54f10b938a03b00bd1d6c Sun, 22 Apr 2012 11:28:32 GMT http://channel9.msdn.com/Forums/Coffeehouse/Machine-Learning/73c0740925e54f10b938a03b00bd1d6c#73c0740925e54f10b938a03b00bd1d6c contextfree` 17 http://channel9.msdn.com/Niners/contextfree`/Discussions/RSS Coffeehouse - Machine Learning

, contextfree` wrote

 should probably sign up now ...

Thats what I thought, I need to find out how much it costs, thanks

]]> http://channel9.msdn.com/Forums/Coffeehouse/Machine-Learning/fe938b74b8364e6eace1a03b00c51b73#fe938b74b8364e6eace1a03b00c51b73 Sun, 22 Apr 2012 11:57:38 GMT http://channel9.msdn.com/Forums/Coffeehouse/Machine-Learning/fe938b74b8364e6eace1a03b00c51b73#fe938b74b8364e6eace1a03b00c51b73 Vesuvius 17 http://channel9.msdn.com/Niners/vesuvius/Discussions/RSS Coffeehouse - Machine Learning @vesuvius: It's a free course (video lectures + auto-graded quizzes and programming assignments) - there is no real college credit or anything associated with it though, so if you're looking for something like that you'd have to look elsewhere, but it's a good way to learn the material I think.

]]> http://channel9.msdn.com/Forums/Coffeehouse/Machine-Learning/21c8ff38097d47a58a72a03b00cf8cab#21c8ff38097d47a58a72a03b00cf8cab Sun, 22 Apr 2012 12:35:39 GMT http://channel9.msdn.com/Forums/Coffeehouse/Machine-Learning/21c8ff38097d47a58a72a03b00cf8cab#21c8ff38097d47a58a72a03b00cf8cab contextfree` 17 http://channel9.msdn.com/Niners/contextfree`/Discussions/RSS Coffeehouse - Machine Learning @contextfree`: Even better, I am not really bothered about college credits, as they no longer mean anything in my career

]]> http://channel9.msdn.com/Forums/Coffeehouse/Machine-Learning/bc4c68a9126647719b7da03b00d47c90#bc4c68a9126647719b7da03b00d47c90 Sun, 22 Apr 2012 12:53:38 GMT http://channel9.msdn.com/Forums/Coffeehouse/Machine-Learning/bc4c68a9126647719b7da03b00d47c90#bc4c68a9126647719b7da03b00d47c90 Vesuvius 17 http://channel9.msdn.com/Niners/vesuvius/Discussions/RSS Coffeehouse - Machine Learning What was that early .NET game where you would write code representing insects and the bugs would duke it out across the internet?

-Josh

]]> http://channel9.msdn.com/Forums/Coffeehouse/Machine-Learning/7d4d6e75814a46ac8c7ca03b00dd1313#7d4d6e75814a46ac8c7ca03b00dd1313 Sun, 22 Apr 2012 13:24:54 GMT http://channel9.msdn.com/Forums/Coffeehouse/Machine-Learning/7d4d6e75814a46ac8c7ca03b00dd1313#7d4d6e75814a46ac8c7ca03b00dd1313 Joshua Ross 17 http://channel9.msdn.com/Niners/Jsoh/Discussions/RSS Coffeehouse - Machine Learning

, JoshRoss wrote

What was that early .NET game where you would write code representing insects and the bugs would duke it out across the internet?

-Josh

Terrarium!  I got a travel mug from MS for that.

]]> http://channel9.msdn.com/Forums/Coffeehouse/Machine-Learning/b21a763d7101485bbfc1a03b00e9180c#b21a763d7101485bbfc1a03b00e9180c Sun, 22 Apr 2012 14:08:40 GMT http://channel9.msdn.com/Forums/Coffeehouse/Machine-Learning/b21a763d7101485bbfc1a03b00e9180c#b21a763d7101485bbfc1a03b00e9180c Richard Anthony Hein 17 http://channel9.msdn.com/Niners/Richard.Hein/Discussions/RSS Coffeehouse - Machine Learning @Richard.Hein: Yes! It looks like there is a Terraium2. I'm going to have to spin that one up. 

EDIT: It doesn't seem to work in Windows8, even after installing the 2.0 version of the .NET framework.

-Josh

]]> http://channel9.msdn.com/Forums/Coffeehouse/Machine-Learning/d282afb6d2a4419ba99ca03b00f74e44#d282afb6d2a4419ba99ca03b00f74e44 Sun, 22 Apr 2012 15:00:24 GMT http://channel9.msdn.com/Forums/Coffeehouse/Machine-Learning/d282afb6d2a4419ba99ca03b00f74e44#d282afb6d2a4419ba99ca03b00f74e44 Joshua Ross 17 http://channel9.msdn.com/Niners/Jsoh/Discussions/RSS