turn on suggestions

Auto-suggest helps you quickly narrow down your search results by suggesting possible matches as you type.

Showing results for

- JMP User Community
- :
- Discussions
- :
- Discussions
- :
- Self Organizing Map formula

Topic Options

- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Get Direct Link
- Email to a Friend
- Report Inappropriate Content

Jul 13, 2017 9:53 AM
(4168 views)

Hello,

I used the SOM feature to identify some clusters in my data. This has been very helpful and interesting. Yet, in order to publish, I need to replicate those results in an open platform like R, or, alternatively, have a reference to the procedure / formulas used by JMP for the SOM. Currently using the SOM included in R we do not get the same results, thus I need to know what is JMP's application of SOM i.e. formula, parameters, sequence of the calculations, etc.

I contacted my local support and they suggested I ask here.

Thanks,

Uriel.

1 ACCEPTED SOLUTION

Accepted Solutions

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Get Direct Link
- Email to a Friend
- Report Inappropriate Content

Hi there,

We do not support R code generation, but we support Python code generation which hopefully would count as an open platform for you to publish your results.

After you create your clusters, publish the results to the Formula Depot and from there choose the "Generate Python code" option.

I attached a few screenshots that illustrate the steps.

Good luck and let us know if you have any questions!

8 REPLIES

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Get Direct Link
- Email to a Friend
- Report Inappropriate Content

A good start would be to read the documentation on the SOM algorithm in the Multivariate Methods guide.

Help==>Books==>Multivariate Methods

Help==>Books==>Multivariate Methods

Jim

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Get Direct Link
- Email to a Friend
- Report Inappropriate Content

Thanks, but I already did that. all that's there are general refences and concepts, nothing concrete that you can do math or stats with.

Uriel.

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Get Direct Link
- Email to a Friend
- Report Inappropriate Content

I now suggest that you go to the source, support@jmp.com to get your question answered.

Jim

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Get Direct Link
- Email to a Friend
- Report Inappropriate Content

https://www.jmp.com/content/dam/jmp/documents/en/support/jmp12/Multivariate-Methods.pdf page 70 seems pretty detailed.

Craige

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Get Direct Link
- Email to a Friend
- Report Inappropriate Content

Jul 15, 2017 10:38 AM
(4101 views)
| Posted in reply to message from Craige_Hales 07/13/2017 09:11 PM

thanks for your suggestion, but, once again, I need the actual formulas, algorithm or scripts of the process being done, not just the description. there are different implementations and the devil is in the details...

Thanks,

Uriel

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Get Direct Link
- Email to a Friend
- Report Inappropriate Content

Hi there,

We do not support R code generation, but we support Python code generation which hopefully would count as an open platform for you to publish your results.

After you create your clusters, publish the results to the Formula Depot and from there choose the "Generate Python code" option.

I attached a few screenshots that illustrate the steps.

Good luck and let us know if you have any questions!

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Get Direct Link
- Email to a Friend
- Report Inappropriate Content

thanks for your reply. For some reason my version of JMP doesn't have that option, all the other options are there, but the "publish cluster formulas" is not. I have pro 13.1 with an academic license.

In any case, that is not what I meant. The option you suggest happens *after* the results are obtained. It just allows for sharing of the results. What we need is to understand how are the results produced. There is an algorithm and/or formulas that the program uses to go from raw data to the clustering results. that is what we need. It is just an implementation of the SOM with K-means. The (very general) description of the process is mentioned, but we need the actual implementation - what is the formula, what are the steps taken? Currently it is a bit like a black box: we give data, select parameters, and then get results, without knowing exactly what the program is doing, other than the general description given. In order to publish we need to be able to document or refer to the actual math being done.

thanks,

Uriel.

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Get Direct Link
- Email to a Friend
- Report Inappropriate Content

The same documentation mentioned above lists the following references for the SOM.

Kohonen, T. (1989). Self-Organization and Associative Memory. 3rd ed. Vol. 8 of Springer Series in Information. Berlin: Springer-Verlag.

Kohonen, T. (1990). “The Self-Organizing Map.” Proceedings of the IEEE 78:1464–1480.

In addition, the kernel used is a Gaussian kernel:

, where *r _{c}* is the location of cluster