[FRIAM] capitalism vs. individualism

[glen]

I'm not sure I agree. Even without unification into a singular whole, we can register novelty by clustering. Clustering in a space, obviously, requires a space of some sort. But spaces are defined by bases that are often only tiny slices/aspects of the things arranged in the space. E.g. we can organize TV shows by run-time, ignoring all other aspects. And if a new show has a run-time different from all other TV shows, then it's novel, even if in an uninteresting way.

I've recently been exploring state space reconstruction methods for some of our more enigmatic model traces. EEMD revealed an interesting IMF for a periodicity I have yet to explain mechanistically. It's a bit infuriating because I wrote the damned model. Anyway, such a task is less about unifying the contributions to the signal than it is finding a basis from which to "debug" it. (Debug in quotes because the periodicity might end up being a counter-intuitive feature.)


On November 5, 2019 12:02:26 PM PST, Marcus Daniels <marcus at snoutfarm.com> wrote:
>Glen writes:
>
>"But re: avoiding modeling the space between the -isms, I'd argue that
>sometimes (only sometimes), it's best to leave the interstitial space
>unmodeled to avoid biasing the integration."
>
>"The best way to predict the future is to invent it."    The space
>unmodeled could contain a configuration (a new Ism) that is has better
>properties than the existing configurations, and the available
>observations are just what has been found so far.    If one is unable
>or unwilling to compress to commonalities -- to unify -- then one cannot anticipate novelty either. I have 500 channels of crap on cable (more, I guess), and I don't really need to watch it all to appreciate the exceptions to this.


-- 
glen