I tried Kohonen for classifying timbres once and though I'm certainly not an expert on it Ben I seem to remember that it works within a threshold value. Input vector is just a _point_ in an n-dim space with a "sphere of influence" around it. If an output vector is within this threshold area then the more input stimulus you present that match a vector falling inside the threshold the more that output is weighted. Eventually that output vector will fire with probability= 1 (absolutely certain). Further input vectors, no matter how many you give, won't affect output vectors outside the threshold. It's a long time ago, so excuse me if I'm talking rubbish.

On Fri, 12 Oct 2007 11:27:15 -0700 "B. Bogart" bbogart@goto10.org wrote:

Hey all, Johannes in particular,

According to Medler: http://neuron-ai.tuke.sk/NCS/VOL1/P3_html/node28.html#SECTION000450000000000...

"...Consequently, more frequently occurring stimuli will be represented by larger areas in the map than infrequently occurring stimuli."

Which leads me to believe if you present a SOM with the same pattern, the number of nodes that fire should increase proportional to the frequency of the pattern.

Problem is I can make ann_som do that, not matter what rules I specify (using the regular decreasing learning rate).

If I send the same pattern to the som 10000 times, only one node will fire for that duration.

Am I misinterpreting the text?

Also by default does ann_som using time-seeded random numbers or regular random numbers? I ask because it appears the same node always wins at the start... (the biggest number).

Is there a FANN mailinglist for the discussion of such issues?

(http://leenissen.dk/fann/ is timing out for me.)

Thanks, B.

---

PD-list@iem.at mailing list UNSUBSCRIBE and account-management -> http://lists.puredata.info/listinfo/pd-list

Quoting "B. Bogart" bbogart@goto10.org:

Hey all, Johannes in particular,

According to Medler: http://neuron-ai.tuke.sk/NCS/VOL1/P3_html/node28.html#SECTION000450000000000...

"...Consequently, more frequently occurring stimuli will be represented by larger areas in the map than infrequently occurring stimuli."

If I send the same pattern to the som 10000 times, only one node will fire for that duration.

Am I misinterpreting the text?

i think so. [ann_som] will only output 1 winner at a time (probably it will output
severeal winners if they all have the same distance; but i doubt
that...) so the neighbours of the original (and consequent) winner will adapt
to the pattern, but they will always be a "tiny bit" less adapted than
the winner (because originally, they were father away) at some point, the finite precision of the numeric representation will
eventually make the nodes _really_ equal.

Also by default does ann_som using time-seeded random numbers or regular random numbers? I ask because it appears the same node always wins at the start... (the biggest number).

hmm, i cannot remember; i don't think that the random numbers are
time-seeded; but most likely you can specify the seed. you could also initialize the nodes with a constant value. i might have a look at it within the next days. for now, just save the SOM after initialization and compare 2 SOMs in
your favourite numeric program....

Is there a FANN mailinglist for the discussion of such issues?

i don't know, georg should be able to help you here. nevertheless, bear in mind that [ann_som] is a standalone
implementation and has nothing to do with FANN (well both are dealing
with ann, obviously but that's it; i didn't know of fann when i wrote
[ann_som])

mg.st IOhannes

This message was sent using IMP, the Internet Messaging Program.