irr  wrote:

> "Tim Tyler"  wrote in message
> news:c5s74s$20p6$1{at}darwin.ediacara.org...
> > IRR  wrote or quoted:
> >
> > > So for example (and in loose reference to one of your other posts),
> there
> > > are many, many groups of bacteria that cooperate, differentiate into a
> > > multitude of morphologically distinct cell types, and build elaborate,
> > > macroscopic (differentiated though monoclonal) structures.  [See, for
> > > examples, hyphae-producing Streptomyces, akinete and/or heterocyst
> forming
> > > cyanobacteria, iron-scavenging Shewanella macrocolonies, or
the wondrous
> > > assortment of organisms that comprise e.g. thermophilic, alkaline
> microbial
> > > mats].  Yet the Kolmogorov complexity of the genomes from these
> organisms
> > > are indistinguishable, they all have roughly the same number of genes,
> the
> > > degree of their metabolic and protein interaction networks
are all scale
> > > free (power-law distributed) with nearly identical scaling exponents,
> etc.
> > > etc. ad nauseum.  Find me some underlying, unifying Standard Model of
> > > Complexity that has evaded notice in all studies to date, and I'll pay
> all
> > > the publication charges on your paper.
> >
> > The best-established way of measuring complexity is to use Kolmogorov
> > complexity.
> >
> > This:
> >
> > * Confines discussion to digital phenomena;
> > * Is difficult to measure;
> > * Has a subjective element - since it depends on a choice of
> >   descriptive language.
> >
> > The first is no problem, if we are content to confine ourselves
> > to the complexity of genomes.
> >
> > The second is a problem in theory:
> >
> > * Except in a few trivial cases, you can only put bounds on the metric -
> >   rather than measure it exactly (and even then the lower bound is rarely
> >   much use).  I would suggest ignoring this problem - and measuring
> >   the value using a conventional high-quality compressor of a type
> >   that is capable of dealing well with repeated sequences.
> >
> > ...and in practice...
> >
> > * You need to sequence the genome in question before you can measure
> >   its complexity;
> >
> > The third makes the metric less asethetically attractive.  My approach
> > would probably be to say something along the lines of:
> >
> > "Always use FORTRAN-77 as your language".
> ....]
> 
> IMO this third problem -- choosing a language with which to quantify
> complexity -- is still *the* showstopper when it comes to biology.  While we
> might all agree that the primate brain is an incredibly complex organ, it's
> not at all agreed upon what it is we mean by this.  For example, a
> Kolmogorov measure fails miserably in classifying the brain as complex,
> after all you're really only talking about two dozen or so different
> recognized cell types stamped out in enormous repetition with iterated
> connections between them -- in other words, a digital representation of the
> brain is incredibly compressible.  I can think of examples with a fraction
> of the genes present in a human brain cell that can give rise to at least
> that many different recognizeable cell types, and I can just as easily think
> of organisms with a vastly larger proteome that show none of the
> interactions or differentiation of neural progenitor cells.
> 
> In my mind the obstacle here is that we have not yet rightly /defined/ the
> problem.  Our inherent notion of what is complex is prominently if not
> exclusively based on (as evident in my examples) visual cues, but prominent
> biological observables such as cell morphology have been -- at least thus
> far -- incredibly poor indicators of complexity.

Chaitin suggests that you pick a suitably comprehensive and capable
language - he uses LISP - and stick with it. So long as you can express
or model what you are trying to comparatively measure, then the results
are notionally commensurate. But the problem here is that this makes
"complexity" a measure of what we can *say* about something. And this
seems right to me - complexity is the surprisal value, as it were, of
our descriptions of things. The relative scales will change as we are
able to describe more and more, unless we artificially restrict
ourselves to some aspect of the things, like the DNA (introns or exons?
Expressed or regulatory? Methylation or not?) sequences of organisms, in
which case complexity is a measure of some theoretical quantity and is
*still* an abstract notion. I don't see a way to avoid measuring only
what we know about things in this regard.
-- 
John Wilkins
john_SPAM{at}wilkins.id.au   http://www.wilkins.id.au
"Men mark it when they hit, but do not mark it when they miss" 
                                               - Francis Bacon
---
þ RIMEGate(tm)/RGXPost V1.14 at BBSWORLD * Info{at}bbsworld.com

---
 * RIMEGate(tm)V10.2áÿ* RelayNet(tm) NNTP Gateway * MoonDog BBS
 * RgateImp.MoonDog.BBS at 4/19/04 7:32:28 AM