Wednesday, March 08, 2006
Generative Methodology Glass Bead Games
On ontological modeling of expression
The metaphor between gene, cell and social
expression à [217]
On Formal verses Natural systems à [206]
Comments by Paul Prueitt
Comment
by Judith Rosen à [219]
Extended
comment by Paul Werbos à [220]
(A letter to a colleague this morning)
In the stratified theory (that we feel is correct), from the perspective of natural systems; natural categorization occurs at a number of organizational levels. These organizational levels are both self emergent and reflective of structural delineation into levels.... as observed in the prominence of so called "pi" numbers; Planks constant and Boltzman's number - being examples. A more detailed description is given in Chapter Two, Foundations of Knowledge Science. Note that natural emergence is a key focus of this theory.
This theory
From this theory about natural phenomenon one can conjecture that evolving natural type shapes the patterns occurring in Internet transactions spaces. More importantly, natural phenomenon is everywhere present in the causes of Internet transaction spaces. Examples: motivation to earn a living in business spaces, motivation to replicate self-identity in gene and cell expression spaces; motivation to win in war, etc. These natural phenomenon are both stable and subject to several types of transformations. The OASIS Business-Centric Methodology captures some of the elements of this theory by making distinctions between “pragmatics” and “semantics”, as well as distinctions between Information Technology expertise and domain knowledge. I have captured these distinctions in a framework using the two figures below.
Simple framework for communicating about some basic distinctions
The phrase "semantic pattern alignment" is used in the OASIS BCM specification (2004), as well as other specifications. Semantic pattern alignment can mean an alignment to technology dependant OWL class definitions, as in BioPAX; or it can mean that a "second school" type activity is occurring that mines well understood "information structure" (such as a BioPAX knowledge base - ie description logic based information model plus lots of data), so that n-ary type ontology (ie outside of OWL and description logic) is produced. (A type of semantic extraction.)
These n-aries can be pruned and joined using fast convolution operators (again without any form of automated theorem proving), [1] to produce simple n-aries having certain inverse properties (Dr Richard Ballard's work). My work with Dr Peter Stephenson on cyber event detection has explored this methodology in the context of recognizing the complex mechanism used in attacks.
The Novelty Detection in Massive Databases proposal, we were almost funded for in 2003, was to demonstrate that the scalability problem becomes, not only sub-linear, but saturates (ie after a initial period the complete set of categories becomes known). Only as a function of time will new categories be introduced.
One of the early n-ary interfaces (2001)
Why would a "properly" reifiable information structure (like one of these n-aries) be "invertible"? What does "invertible" mean? I see this in a certain way, and Ballard sees this in a slightly different way. What I am doing now is to work within what the 2004 OASIS Business-Centric Methodology refers to as the pragmatic axis, where the uniqueness of a situation is acknowledged. Thus this work addresses the recognition of novelty (which was about 50% of the focus of my 1988 PhD in applied mathematics.)
In the theory I developed, cA (categoricalAbstraction), the "semantic axis" is categorically stable, whereas the pragmatic axis is not. The stability is critically important to B-2-B transaction analysis. The inversion of "correct" semantics should reflect an invariance that is related to the semantic axis but not the pragmatic axis.
At least this seems to give language to the distinction that not all aspects of a transformation are formalizable as description logic reified classes.
BUT, a certain part of a real transaction can be recognized using semantic pattern alignment methods (for example as proposed BCM templates). The other part should, induce a measurement that, acts as a residue that informs an actionable intelligence cycle of evidence that residue (unexplained part of the measurement) exists. Gauge theory comes to mind.
The cancer Biomedical Grid project is likely following the example, or moving in a similar path, as the BioPAX OWL ontology. The issues related to distributed knowledge representation within a community of medical scientists are pretty well exposed in several bioinformatics projects involving OWL. My group is following this work closely.
What we see is the standardization of some subset of the concepts that are needed to conduct science about biology.
But there are important open questions. How natural categories form in vivo (in reality) has not been reflected in how classes, properties and restrictions reify instances, as yet. Perhaps we will continue to make progress in this direction. What are the mechanism(s) in which natural interaction systems preserve self (what is the replicator function)? How does reconciliation occur, or not occur in these transaction spaces? What might be reasonable reconciliation methodology? Why does “what”, and why does “how” occur in vivo the way (we observe) them to occur?
In discussions within a small community, talking about Robert Rosen's work on natural entailment patterns, my group is finding one possible approach towards a cyclic measurement activity that is largely automated. This measurement methodology uses something like gauge theory as a means to automate the transformation of instance data into ontological classes. The buzz-phrase is the “compression of massive amounts of data into information structure”.
What my group’s work contemplates is the (non-statistical) identification of "all" invariance in a massive data set and representation of this invariance with a set of n-aries as discussed in the Notational Paper:
http://www.bcngroup.org/area2/KSF/Notation/notation.htm
and then the presentation of "semantically" underconstrained symbols for differential interpretation by humans (or computer programs using stochastic induction). This is referred to as differential and formative ontology, in our work.
If only I knew which funding programs might be open to this type of approach, I am sure that funding would flow and support this community in their basic scientific effort. But I am too far outside the funding loop to even care anymore.
It sounds like your proposal addresses issues close to this.
Categorical abstraction and event Chemistry are terms I invented when working on the cyber event detection domain in 2000.
http://www.ontologystream.com/area2/REAL/SIL&G.htm
The link above was focused (2003) on two outcomes.
One was the development of a community owned set of patents that would permanently fund a radical new approach to observing natural category formation and evoking a human synthesis of new symbols with correspondences to the observations. The organizational structure is proposed to be similar to that of the Behavioral Computational Neuroscience Group (www.bcngroup.org ) charter.
This organizational proposal, regarding patents, was likely a mistake, as this proposal was accepted as a final cut for funding (deemed fundable) but not funded. I understand that at the last moment Cyc Corp founder Doug Lenat made a personal appeal for the funding that SAIC/Ontologystream would have received (in late 2003).
Needless to say, this was one of seven major proposals made and almost funded in 2000 to the present; and the absence of funding has been a fact of life for members of the group.
At the present time, my group stays on topic pretty much as a hobby; and through dedication to what is life long effort in scholarship.
Anyway, I did not mean to write such a personal letter. I hope that NIH accepts your approach and that your funding occurs.
Dr Paul