From: Walter Hicks (wallyshoes@mindspring.com)
Date: Sat Nov 01 2003 - 08:33:24 EST
Dawsonzhu@aol.com wrote:
> Walter Hicks wrote:
>
>
>> I can say as others have said to me: If you
>> want to understand GA, you need to get a book
>> and read it.
>>
>
> Fair enough, so why not suggest a leading book
> on the subject so
> we can all decide for ourselves if there is
> something significant in
> this or not.
Hi Wayne,
GA is a method that applies what is the output
(models) from your area of expertise to other
disciplines. Which book is good is a function of
what you intend to use it for. In your case, it is
evolution itself (I think). In that case, your
following text agrees with the subject line of
this post. (i.e. evolutionary theory does not yet
work.) It is a "work in progress" and that is fine
with me.
I have been involved with GA being applied to one
specific area. In one situation, where the problem
was simple and the answer was known, GA converged
to the correct answer. In more complicated
situations, where multiple approximate answers
were available, it did only moderately well and
other approaches had to finish the job.
For a GA reference in a given field, I would say:
1.) All God's children (on this list) have Google.
2.) Amazon.com probably has every book (913)
available on the subject -- chose your field.
Walt
>
>
> Working from a different angle of computer
> programming, I notice
> that a common approach I use is to start with a
> working program and
> modify it to make it to work for another
> project. I will also combine
> different programs together to build a new one.
> As I build more related
> programs, I have more tools to exchange between
> them, and a kind of
> natural evolution of the program occurs. I also
> sometimes have
> subprograms I leave decommissioned, only to
> discover that it can be
> re-invoke with minor modification at some
> point. I also might switch
> between decommissioned subprograms, and the
> newer one for various
> reasons. Within all of this, there are
> fundamental parts of the program
> that simply do not change or change only in the
> most superficial ways.
>
> So these have remote analogies with simple
> mutations, horizontal
> transfer, neutral mutations, and pseudogenes.
>
> As long as the modifications are reasonably
> small, that strategy works,
> so given one has the machinery one needs, it's
> not hard to work toward
> a goal. But there are times when I find I
> really have to rethink the approach
> because the old way just cannot meet the
> requirements. It is not the
> efficiency issues in my case (which is a whole
> different matter from just
> doing a job), it is a matter of the fundamental
> needs of the task. In such
> cases, I may have to do a major rewrite and the
> result is something that
> has little resemblance to the original, except
> for possibly some retained
> variables. When it can be modified step by step
> toward that ultimate goal,
> it is again a clear cases of a selection
> process, but sometimes I must make
> the whole modification before I even have the
> chance to test it. It is not
> just an "enough time" issue, it is really
> building a whole module from scratch.
> It is also not a mere issue that I have built
> extra stuff that could have found
> an easier way to the goal, there is no way to
> the goal except by building that
> new module from scratch. I do not like to waste
> time writing major code from
> scratch unless I have to.
>
> If we introduce efficiency issues, that would
> make the matter worse because
> special algorithms designed for speed are less
> amenable to major modification.
>
> So in short, I can grant you some room there.
>
> What I can realize from this comparison is
>
> (1) Extinction is most likely to occur when the
> rate of adaptation required
> is faster than the intrinsic capacity of an
> organism to make that adaptation.
>
> (2) It leaves me with some unresolved questions
> about where new functional
> genes come from.
>
> (3) In connection to (2), I must ask what level
> of complexity is required for
> any given adaptation in the history of the
> earth.
>
> (4) I think introducing RNA/DNA is of minor
> significance, it helps preserve
> neutral mutations in functional genes, but aside
> from physical issues of
> biopolymers typically possessing a persistence
> length of the order of 3
> nucleotides (persistence is a measure of the
> bendability of the polymer), it
> would be just as efficient if there was one
> single index. So aside from
> ribosomal RNA being a living fossil, I would
> dismiss that as a relevant argument.
>
>
> Point (3) is what requires some unpacking. What
> I am asking is
>
> (a) Could we start with some core fundamental
> machinery
> (both RNA/DNA and protein components) and build
> all of life with it?
> This is my opinion, but I don't really think a
> _pure_ RNA world is feasible.
> There must have been some mutual co-development
> of these two
> systems.
>
> (b) Can we trace our way back to these core
> pieces of machinery?
>
> (c) Are these core pieces of machinery
> sufficiently simple that such
> units could occur in a hydrothermal vent or some
> other proposed location
> of natural development? And, are they enough
> that all other functional
> proteins could (in principle) develop? I think
> we should not ignore the
> challenge of _how_ to get a piece of ribosomal
> RNA and gene replication
> machinery started in the first place and this is
> something that molecular
> evolution often seems to ignores taking it as
> "given". We have the low
> end in the pea soup of life, and _given_ rRNA
> etc, we _may_ have the high
> end, but it is hard to see a lot of simple steps
> in-between.
>
> This is not a defense intelligent design or
> irreducible complexity,
> I think these are largely unanswered questions.
> We may find
> answers to them, but most of them are only
> currently partially
> resolved.
>
> At any rate, a genetic algorithm (GA) can only
> help if the components put
> into the pool are sufficient for the job. If
> the programmer didn't give
> the GA enough to do the job in the first place,
> the programmer cannot
> expect a stupid GA to help him/her do the work
> that he/she was supposed
> to do himself/herself. It would be nice to have
> a computer like on Star Trek
> where I just order the computer to do some
> arbitrary job and it simply does
> following my _intended_ request and not what I
> asked it to do. Wow, what
> a remarkable machine that is!
>
> It does put some onus on people who simply take
> as given that some very-
> very-very powerful machines are just simply
> there (dropped down from the
> sky I suppose). But, at the same time, where we
> draw the line on what is
> feasible remains unanswered also. I addition,
> even when given, what is
> sufficient is yet an unanswered question.
>
> by Grace alone we proceed.
> Wayne
-- =================================== Walt Hicks <wallyshoes@mindspring.com>In any consistent theory, there must exist true but not provable statements. (Godel's Theorem)
You can only find the truth with logic If you have already found the truth without it. (G.K. Chesterton) ===================================
This archive was generated by hypermail 2.1.4 : Sat Nov 01 2003 - 08:36:57 EST