Altair RISE

A program to recognize and reward our most engaged community members

Nominate Yourself Now!

"SOM (SOMDimensionality Reduction) documentation or examples"

Hi all,
I'm looking into the use of RMiner's SOM. I can get it to produce nice looking results but would like more information on its setup and interpreting the type of results in RMiner.

Does anyone have any worked examples or documentation? I'm wondering, in particular, how to interpret the resulting plot view and data view, relating the points back to the original data features.

I'd be very grateful for any help.

Thanks,
Richie

Find more posts tagged with

AI Studio

Text Mining + NLP

Accepted answers

All comments

haddock

Hi Richie,

I think I told you in a previous post that RM is much better than its documentation, this is a case in point. Try this code, and waft your mouse over the data points of your SOM plot ( coloured by label probably ), behold the Ids!

<operator name="Root" class="Process" expanded="yes">
    <operator name="ExampleSetGenerator" class="ExampleSetGenerator">
        <parameter key="target_function"	value="random"/>
    </operator>
    <operator name="IOMultiplier" class="IOMultiplier">
        <parameter key="io_object"	value="ExampleSet"/>
    </operator>
    <operator name="IdTagging" class="IdTagging">
        <parameter key="create_nominal_ids"	value="true"/>
    </operator>
    <operator name="SOMDimensionalityReduction" class="SOMDimensionalityReduction">
    </operator>
</operator>

While the human eyeball mk.IV is about as smart as it gets in terms of finding and enjoying patterns, it sometimes may like patterns that are not helpful; in this latter case wheel out some horrible counting stuff to give the semblance of reason. In your context it is a matter of checking that the reduction produces "better" performance. So cross validate a learner on the original dataset, then reduce it and repeat the process. If it does, even after you've messed around with genetic optimisers etc.etc.., then you may have found an abstraction.

;D

I stress the phrase "semblance of reason", because no matter what you do you will be left with doubt, known here as the "Curse of MIerswa" - did you realise that "Ingo Mierswa" can be rearranged as "imagine rows"?

dubito ergo sum ("I doubt, therefore I exist") and si fallor sum ("If I am deceived, I exist")

Augustine of Hippo

Ergo clearly a dataminer.

IngoRM

I stress the phrase "semblance of reason", because no matter what you do you will be left with doubt, known here as the "Curse of MIerswa" - did you realise that "Ingo Mierswa" can be rearranged as "imagine rows"?

Amazing! I tried so many years to keep my true nature hidden as deep as possible. Like other suspicous creatures I of course was not able to bear this burden all alone and started to give hints to others. I have to admit that writing an application (RapidMiner) to give others a hint to the true purpose of myself was maybe too strong a hint but nevertheless: after only eight years of development, finally somebody (you) managed to break this secret and unveil my secret nature.

Of course, I am sure that you noticed that "Ingo Mierswa" is also an anagram of "Image In Rows" (Image, Pattern, all the same) and of course "Mirage Is Now" which I personally like most

By the way: thanks again for your valuable posts which I enjoy not only reading because of their quality in terms of the topic but also because of the quality of language. You really should consider to write a true "Data Mining Novelette". I would certainly buy it!

Cheers,
Ingo

IngoRM

By the way: if you rearrange "RapidMiner" you will end up with "Repair Mind". More thorough investigation might be necessary here.

Cheers,
Ingo

haddock

Excellent
Bravo, bravo ;D ;D ;D

steffen

Ingo Mierswa wrote:

By the way: thanks again for your valuable posts which I enjoy not only reading because of their quality in terms of the topic but also because of the quality of language. You really should consider to write a true "Data Mining Novelette". I would certainly buy it!

I totally agreee ! I also enjoy reading the valuable suggestions and the fine humour of Mr Haddock