computationally evolving cell development...

Brian Foley brianf at med.uvm.edu
Tue Aug 23 11:00:24 EST 1994


Tony Hirst (A.J.Hirst at uk.ac.open) wrote:
: Be forewarned - I'm a newcomer to this group on an info.quest with a
: beginner's question... 

: I'm just about to start looking at ways of using genetic algorithms to
: implement a developmental cycle in a parallel controller for a mobile
: robot, and in a fit of scrabbling for biological info. I'd appreciate brief
: answers to the following questions (if poss. - i appreciate they may be
: MOL-type questions...!?!)...

: My understanding of DNA is that it codes for proteins(?) that are built via
: mRNA working plans and tRNA builders, in the building site that is the
: cytoplasm(?)...where do the building blocks come from????? 

More biology questions deleted....

: - by mail, if you don't wnat to clutter this list

I replied with a long letter by mail.

: - by posting, if you think it's of interest/use to the group 

I want to point out here that genetic algorithms have no need to
be similar to biology, other than the (duplication with "mutation" 
followed by selection) repetitive process.

It seems that most biological evolution occurs by "random point
mutations" with a significant amount of "insertion/deletion/recombination"
also contributing.  But none of it follows rational rules.  The cell
does not design new mutations, nor does the organism (exluding humans
which are just beginning to think about doing so).

Genetic algorithms need not make mutations entirely at random just
because that is how it seems to work in biology.  In GAs we can
create ratioanl schemes for altering the "genome" at each cycle,
rather than just doing random bit flips.

If you wanted to use a genetic algoythm to write a novel, you would 
need billions of CPU hours to do it by randomly inserting ASCII characters
into a set of files and selecting the one that looks most like a
book at each round.  It would be far better to tell the mutation
part of the code something about words and sentance structure, so that
mutations rearanged whole words rather than single letters.  Later in
the evolution, perhaps whole paragraphs would be moved, rather than just
words.

: thanks in advance,

: monty


: --------------------------------------------------------------------------
:  All opinions etc etc...
: --------------------------------------------------------------------------
:       | Tony Hirst ("Monty")          | e-mail:  A.J.Hirst at open.ac.uk
:       |                               |
:       | Human Cognition Research Lab. | phone:   +44 (908) 65-4404
:       | The Open University           |                    65-4481
:       | Milton Keynes   MK6 6AA       | 
:       | BRITAIN                       | 
:        -------------------------------------------------------------------
:       | "There is no meaning..."         "Science is a subset of art..."
:        -------------------------------------------------------------------

--
********************************************************************
*  Brian Foley               *     If we knew what we were doing   *
*  Molecular Genetics Dept.  *     it wouldn't be called research  *
*  University of Vermont     *                                     *
********************************************************************



More information about the Mol-evol mailing list