Message Boards Message Boards

0
|
5602 Views
|
6 Replies
|
0 Total Likes
View groups...
Share
Share this post:

use fuzzy C-means

Posted 9 years ago

I wand to use Fuzzy c-means to give clusters of image like function ClusteringComponents. But ClusteringComponents don't use fuzzy logic.How can I don this? I don't find in manual anything about Fuzzy C-means. But I found that in old versions Fuzzy logic is additional packet. And I shoud buy and add this to mathematica by myself. But I read, that there is Fuzze c-means in Wolfram Mathematica 10. How can I use Fuzzy C-means in Wolfram Mathematica and where I can read any information about fuzzy functions with examples because I didn't find information in help (F1 button). Thank you

POSTED BY: Dmitry Dmitriev
6 Replies

Fuzzy C-Means is defined in the Fuzzy Logic package. Unfortunately, this package is old and outdated. I'm not sure it will work in Mathematica 10. It seems to have last been updated for version 5 of Mathematica.

But I read, that there is Fuzze c-means in Wolfram Mathematica 10.

Where did you read this?

Most likely, this is something worth just programming yourself. Alternatively, you could use a package in the R langauge do this and then use R-Link, which allows you to call R code from within Mathematica.

POSTED BY: Sean Clarke

I found fuzzy C-means code in matlab and in C#. Can I import functions from other languages to Wolfram? How can I do this?

POSTED BY: Dmitry Dmitriev

Yes, you can use .NET languages like C# using .NETLink:

http://reference.wolfram.com/language/NETLink/tutorial/Overview.html

This is most likely not the solution you want though. My reccomendation is still the same. I would write out the routine in Mathematica. Or I would use RLink to use a R package to do this.

POSTED BY: Sean Clarke

Can you give me any more information about R-link and R package. I don't understand what is this. Also I have one same question. How can I use neural networks in Wolfram?

POSTED BY: Dmitry Dmitriev

I haven't used RLink in a while. I would first go through the RLink tutorial:

http://reference.wolfram.com/language/RLink/tutorial/UsingRLink.html

This documentation is available in Mathematica as an evaluatable notebook.

I didn't have a particular R package in mind, but if you search you can find implementations of this.

http://web.mit.edu/r_v3.0.1/lib/R/library/e1071/html/cmeans.html

https://en.wikibooks.org/wiki/DataMiningAlgorithmsInR/Clustering/FuzzyClustering-FuzzyC-means

POSTED BY: Sean Clarke

When asking new questions, I would reccomend creating a new thread. This makes sure that other people can see the new question. Sometimes the person who answered your first question isn't the best person to answer the other question.

You can use the Classify function in Mathematica. This doesn't give you any control over the neural network, but allows you to create and use a simple FF for classification.

There was an add-on package for neural networks. It's a bit old and wasn't updated for the latest version of Mathematica. I've been told it works fine for the most part with Mathematica 10. There has a been a significant interest by the devleopers to incorporate it's features into a future version of Mathematica.

I've written once or twice some toy, simple neural networks. Mathematica has enough built in funtionality to make writing your own lightweight neural networks pretty easy, but you have to be familiar with the language.

Searching online, I see this package, but I'm not familiar with it: http://evolution.felk.cvut.cz/recurrent/

POSTED BY: Sean Clarke
Reply to this discussion
Community posts can be styled and formatted using the Markdown syntax.
Reply Preview
Attachments
Remove
or Discard

Group Abstract Group Abstract