Re: Re: Feature Engineeriing!
> -----Original Message-----
> From: owner-standard-upper-ontology@LISTSERV.IEEE.ORG
> [mailto:owner-standard-upper-ontology@LISTSERV.IEEE.ORG]On Behalf Of
> Rich Cooper
> Sent: Monday, 7 March 2005 2:13 PM
> To: Rob Freeman; John F. Sowa
> Cc: 'SUO WG'; cg@CS.UAH.EDU
> Subject: Re: Feature Engineeriing!
>
>
> > The mistake is the same in both cases. The solution is not better
> > features,
> > but to abandon the assumption those features organize themselves in only
> > one
> > way.
> >
> > -Rob
>
> Your approach is very interesting Rob, and the one about
> retaining many meanings in the natural language representation
> rings a strong bell with me. I wish I knew how to take the next
> step and find a representation that preserves the many levels
> of interpretation that people place on a phrase.
>
Done (or more so the general processes identified ;-)). From my research on meaning in general we find that:
Given a *finite* set of POSSIBLE qualities used to derive meaning, so ALL of the qualities are reflected THROUGH each of them. This is akin to DNA encoding of one cell containing the whole genome. (in DNA codon generation the recursion is of purines/pyramidines)
Recursion of WHAT/WHERE aka differentiating/integrating means that as we recurse 'down' so we use each derived row as a 'language' source. e.g. I can have a row of 4096 'qualities' serving as UNIVERSALS, archetypes and so lacking 'colour', that are then in need of context to 'ground' them in the LOCAL. (see diagram in http://www.iimetro.com.au/~lofting/myweb/btree.gif) [note that Peano used recursion of the empty set to 'formally' derive integers - this is akin to the open-ended nature of the recursion 'down' of a dichotomy of A/NOT-A as compared to the CLOSED nature of the rows]
What is also noteworthy is that the universals, if not associated with some local context can develop within their own - e.g. PURE mathematics (and so notions of '2') vs APPLIED mathematics (and so notions of '2 of something').
We have the apparent hard-coded nature of generic meanings, or more so brain oscillations that reflect recursion of general to derive increasingly particular - thus the use of the attention system to focus and so particularise, combined with the oscillations WITHIN those bounds, allows for the generation of meaning. The oscillation is across the ASYMMETRY in the brain, and that asymmetry is a dichotomy of differentiating/integrating, aka XOR/AND etc.
The HIERARCHY that develops allows for each AND state to be XOR-ed where that process derives parts - the easiest example being in the processing of sensory paradox (http://www.iimetro.com.au/~lofting/myweb/paradox.html)
What the research shows is that given a TEMPLATE of meaning derived from recursion of a dichotomy so we overlay that template with particular dichotomies. The template structure is such that it, when represented as recursion of 1/0, applying XOR of quality A to quality B will give the A-ness of B described by analogy to some other quality C.
What this process shows is that, given that A, B, and C are derived from the same row, so meaning is derived by analogy due to the finite state of the set of POSSIBLE meanings given a universe of potentially infinite meanings - this takes us into the differences of NOUNS/VERBS vs all possible labels for nouns/verbs. The brain has derived ONE generic format for meaning derivation that is then relabelled for each unique context to which it is associated.
Given that all we can know is based on recursion of differentiating/integrating (and you can add mediation to map to the encoding of CONSCIOUSNESS in the dynamics) so we can recurse the dichotomy ad infinitum and generate row after row of finer and finer discrete 'languages'.
Thus, it is not the recursion of dichotomies, trichotomies etc etc that is of interest, it is what happens when we XOR each derived quality with all of its 'kin' - we find a source of meaning far beyond the basic categories derived in each row - we identify the 'entanglement' of all parts where each row is a finer 'cut' of the whole.
As an abstract example, given recursion of 1/0 three times gives us 2^3 categories:
000, 001, 010, 011, 100, 101, 110, 111
My work shows that these can represent qualities, and XOR-ing them with each other draws out finer details. Thus 010 XOR 110 gives us 100 as the description by analogy of the 010-ness of 110, just as 110 XOR 010 represents the 110-ness of 010. Thus from eight categories we derive 64 differences rather than just eight. IOW the set of qualities takes on a 'sequence' within EACH quality, derived from XOR-ing from first bit through to all bits.
I can map these to basic types of categorisations of personalities, or types of emotions or 'yin/yang' qualities, or to qualities of numbers used in Mathematics (VERY vague but still 'meaningful').
Given a set of qualities derived from a ROW of recursion, so when things get too 'intense', when the qualities in the row are over-used and things get 'complex'. so we can apply recursion generate a new row - akin to 'adding a dimension'.
That said, it appears more that hyperbolic development is used where the WHOLE is used to flesh out each PART. For example, from a set of 8, rather than add a dimension and move to 16, we move to a set of 8 x 8 (eight octets of 'variations on a theme') that THEN get labelled to reflect 64 'unique' qualities. This process allows for us to get a LOT more space to work in and so in the long run CONSERVES energy when compared to regular 'adding a dimension' methods.
Chris.
This message was sent through MyMail http://www.mymail.com.au