machine Discussions Rss Feed

New Post: How to do this ?

SethJuarez — Mon, 04 Mar 2013 20:44:55 GMT

Great question, there are a couple of things you might be able to do. Send me an email so we can start to look at your data. Also, this project has been renamed and moved. To learn more see http://numl.net

New Post: How to do this ?

AdamLoh — Wed, 27 Feb 2013 04:07:35 GMT

Hi Admin,

Thanks for creating machine for .net. Awesome library !!

Let me tell you what i want to achieve.

I have the customer, the products they bought over a period of time, let say 1 year.

I want the system that that will suggest the customer to bought that is similar to their previous product they bought, so that our sales associate will recommend those product as well.

New Post: Different methods of Random generator in Visual C#

suresh__trader — Wed, 23 May 2012 12:02:43 GMT

Dear Sir,In mathematica Software there are different methods are available for creating random number generators as below
1 "Congruential" (linear congruential generator (low-quality randomness))2 "ExtendedCA"(extended cellular automaton generator (default))3 "Legacy"4 "MersenneTwister"(Mersenne Twister shift register generator)5 "MKL" (Intel MKL generator (Intel-based systems))6 "Rule30CA"(Wolfram rule 30 generator)
I need to have above randum number generators in visual c# 2008 and above .Please let me know , Is there any library already exists ?or you can create it ?Thanks.
I am eagerly waiting for your reply.
Regards
Suresh.

New Post: Is it Possible to Interact Directly With the ML Algorithms without Wrapping an Object Around Your Data?

sethjuarez — Tue, 22 May 2012 18:13:00 GMT

I am finally doing this. I completely agree.

New Post: Support Vector Machines

sethjuarez — Tue, 22 May 2012 18:12:30 GMT

Yes! It is definitely one of the things I want to add. Hopefully EOY.

New Post: Support Vector Machines

tbischel — Wed, 29 Feb 2012 18:28:00 GMT

This library is pretty cool... but I feel like the big missing piece of the puzzle are support vector machines. I know the math is daunting, but feel this is the future of machine learning with high levels of accuracy. Are you still developing on this project, and any chance you'd get to SVM's any time in the next year?

New Post: Could Tests Be Written in Something Other Than MSTest

pnewhook — Tue, 16 Aug 2011 13:13:56 GMT

I've migrated the tests in my fork (https://hg01.codeplex.com/forks/pnewhook/mbunittestconversion) but I can't yet open the solution in Express because of solution folders. I'm still working on that.

New Post: Could Tests Be Written in Something Other Than MSTest

pnewhook — Sat, 06 Aug 2011 05:30:07 GMT

Sounds like a plan. I've started my fork and I've migrated CalculusTests and DecisionTreeTests. I think I will need to do some house work before it's mergeable though, I think my mercurial plugin affected the csproj files and mbunit.dll isn't included correctly. On a related note, Test_Decision_Tree_Invalid_Enum_Value failed because it didn't throw an exception, but it also fails in MSTest. I can't figure out why it's supposed to throw and exception though.

New Post: Could Tests Be Written in Something Other Than MSTest

sethjuarez — Fri, 05 Aug 2011 15:31:08 GMT

Cool. I would love for you to convert them. Can you fork the tests and put it into the platform of your choice? I would happily pull in the changes. Make sure you submit the pull request. Thanks!

New Post: Could Tests Be Written in Something Other Than MSTest

pnewhook — Fri, 29 Jul 2011 03:48:53 GMT

I'm not religiously dedicated to any single tool but I think MbUnit is a good fit. It's attributes are very similar to MSTest, which makes migration easy. NUnit is very common, and supported by MonoDevelop, but it's a little behind the curve in terms of feature support (ordered tests via method dependency come to mind). Its test runner supports multiple unit testing frameworks as well, which again makes migration easy. Xunit.Net has some very interesting features, but it's a pretty serious departure from typical .Net unit testing frameworks. A good comparison is found here http://xunit.codeplex.com/wikipage?title=Comparisons&referringTitle=Home

Keep in mind however that you give up really tight Visual Studio integration. TestDriven.Net is a free plugin to non-Express versions of Visual Studio that integrates third part unit testing frameworks, but that still doesn't help the Express and Mono users. I don't know if a CodeRush works better with one framework over another. I realize asking you to give up Visual Studio tooling support isn't trivial.

New Post: Is it Possible to Interact Directly With the ML Algorithms without Wrapping an Object Around Your Data?

pnewhook — Tue, 26 Jul 2011 20:12:40 GMT

I'm not sure I follow your comment about the fixed number of features. I wouldn't need to change the dimensions after I start training, I just can't tell what the features are going to be ahead of time. Is there a way to create dynamic objects?

I'd love to see an example of passing a matrix without first converting from an object.

New Post: Analyzing Text

sethjuarez — Tue, 26 Jul 2011 19:06:57 GMT

Have you seen the StringFeature attribute?

Uh oh! Why are you computing the inverse? What is the error message?

I am using Gauss-Jordan elimination with partial pivoting (which can become unstable with certain matrices)

New Post: Is it Possible to Interact Directly With the ML Algorithms without Wrapping an Object Around Your Data?

sethjuarez — Tue, 26 Jul 2011 19:04:08 GMT

The challenge here is that these algorithms require a fixed number of features when training. I will think about this...

You can definitely just pass vectors and matrices to these things without the explicit object->vector conversion. We should chat!

New Post: Could Tests Be Written in Something Other Than MSTest

sethjuarez — Tue, 26 Jul 2011 19:02:45 GMT

Yessir! Which framework should I use?

New Post: Is it Possible to Interact Directly With the ML Algorithms without Wrapping an Object Around Your Data?

pnewhook — Mon, 25 Jul 2011 17:25:51 GMT

Is there a way to pass a 2 dimensional array instead of a set of objects to build my model? I'd like to create an excel plug-in where the user can pass in a range of cells to train a model, as well as specify a second range that needs to be categorized/regressed/clustered, but I can't tell how many features there are per observation. That makes it difficult to create a container object ahead of time.

Thanks

New Post: Could Tests Be Written in Something Other Than MSTest

pnewhook — Wed, 20 Jul 2011 17:39:59 GMT

Using MSTest for unit testing makes it difficult for those of us using an Express edition or mono. Is there any possibility that tests could be written in another framework?

New Post: What differs this project from others?

rcardare — Mon, 30 May 2011 16:56:29 GMT

Yes, I was wondering the same. I just watched the video from dnrtv.com and that was a question I had. I am also curious how this would scale with large datasets. I use encog but always good seeing different approaches. Thank you for the good list of resources by the way.

New Post: What differs this project from others?

sethjuarez — Fri, 06 May 2011 16:19:51 GMT

Thanks for the great question! The main difference is the approach. My library is specifically designed to do learning off of objects. An implicit conversion is done in order to make the learning accessible to anyone who understands object oriented programming.

Hope this helps!

New Post: What differs this project from others?

asdandRizzo — Fri, 06 May 2011 12:15:22 GMT

Dear Sir,

What is the reason of this project? May be you should help other project?

Or to http://www.cleveralgorithms.com/ by implementing these on C#?

There are some.

On .NET open source:

http://accord-net.origo.ethz.ch/

http://www.aforgenet.com/

http://www.matthewajohnson.org/software/svm.html

http://www.heatonresearch.com/encog

http://sourceforge.net/projects/neurondotnet/

Properitary:

http://research.microsoft.com/en-us/um/cambridge/projects/infernet/default.aspx

http://research.microsoft.com/en-us/projects/sigma/

Wrappers:

http://pytools.codeplex.com/wikipage?title=NumPy%20and%20SciPy%20for%20.Net

http://leenissen.dk/fann/wp/

http://www.emgu.com/wiki/index.php/Tutorial#Machine_Learning_Examples

Machine Learning builded on top of http://en.wikipedia.org/wiki/List_of_numerical_libraries#.NET_Framework_Languages_C.23.2C_F.23_and_VB.NET

Java Machine Learning can be used through http://www.ikvm.net/

Thanks

New Post: Analyzing Text

VanBierStein — Mon, 11 Apr 2011 23:57:20 GMT

Seth, I'm on to a new project where I am analyzing text and there is a numeric value that rate's that texts overall relavency to a peticular subject on a scale from 1-100. In your humble opinion what would be the best method to analyze the text and predict the value for future postings? It is limited to 255 charaters. Right now I am using the linear regression model but wanted to get your opinion on text analysis.

Also I keep running into an unstable matrix on the inverse operation. Is this because the sample data is to small?

Nate