[FRIAM] tools, trollers, and language
glen ☣
gepropella at gmail.com
Mon Jun 12 16:35:30 EDT 2017
Right. My only point was to distinguish the two procedures for examining a thing, because one's choice of procedure can bias one's results. (obviously) With EricS' very detailed throwdown in favor of hierarchical accumulation AND Russ' chosen _target_ of urban systems, I think it's critical that we choose analysis procedures that are as agnostic as possible.
We've now discussed cognitive biases toward _direction_ (up vs. down) and continuity (or population density - laminar flow - AND space vs. graph) ... even if it has taken us days and billions of emails. Are there other biases we could eliminate? I like, but reject, Roger's assertion that "[deep neural nets] don't care about no stinking layers". As with using polar coordinates on an onion (or monotonic "time" in Diffusion Limited Aggregation), deep learning requires at least a sequencing of (distinct) procedures. So, it does require layers in very much the same sense as a DLA. On the other hand, I like considering deep learning as a thing to be analyzed, because it does allow cycles of a kind.
And again, I'm not proposing any of these _things_ are analogs/metaphors targeting "complex systems". I'm only trying to argue for agnostic analysis tools.
TANSTAAFL!
On 06/12/2017 12:45 PM, Steven A Smith wrote:
> At the risk of another discursion:
>
> I think I just realized what I've been (almost) seeing of value in all this back and forth:
>
> 1. I (and Nick) heard Glen's invocation of the Onion as an attempt to
> explicate a useful difference between levels and layers in the
> understanding of Complexity Babble (Talk/Science/Math/???). I think
> he meant only to try to distinguish the two from one another and
> explicate their differences irrespective of the near dead horse we
> were working over at the time. I think this might be the totality
> of the misunderstanding.
> 2. I'm always looking for form/function dualities. In the onion, the
> form (layers) follows a certain functional/behavioural path
> (cyclical growth). I don't even know how to find "levels" in the a
> *hierarchical* sense or otherwise in an onion... maybe if we look at
> the cross section (as Glen suggested) and see *strata* (from the
> source (domain) of geological deposition and erosive or shearing
> exposure?) and then consider drilling a mine shaft into said strata
> which is more suggestive of the term "levels"?
--
☣ glen
More information about the Friam
mailing list