One possible way of extracting the essence of a system is to use it for a while, go away from it long enough to forget some of the details, and then try to duplicate what you've remembered as important. Do you end up with only "essentials" or do you end up with just the "memorable details"?
With c2.com down due to local weather, I attempted this exercise with the WikiWikiWeb server itself. My result was about 400 lines of perl code (17 subs) that is a standalone http server implementing something wiki-like. The results of the above conjecture are not yet clear... -- MarkEichin
Ooh, can you post it ?
I've once heard it said that "experience is what is remembered after all the details have been forgotten".
Exactly the same is done with neural nets:
If the net has too many neurons, it starts learning the training data "by rote" instead of extracting useful generalizations. To counter this, the number of neurons and/or connections between them can be cut down ("pruned"), until the net only remembers the useful abstractions, instead of the details/noise in the training data. --FalkBruegmann
Perhaps this is why people get philosophical when they drink? :-) -- MichaelFeathers
Or when they get old. (No offense to anyone older than me.) BobFarrington
Hypertext documents can be studied as physical objects, and the laws of their behavior ascertained
Critical states of a system are signaled by a power-law distribution
One of the signature features of critical phenomena is a loss of scale at the critical point.