From: Mel Bartels 
To: atm{at}shore.net
Reply-To: Mel Bartels 


> In the Feb 2003 issue of Scientific American  the article of Evolving
> Inventions
> By John R. Koza, Martin A. Keane and Matthew J. Streeter presents an
> interesting concept of design by computer. Optimization programs found in
> some ray tracing software and electrical simulation (spice) work only on a
> range of values of a predefined optical or electrical design. The new type
> of software presented in this article actual mimics the process of
> evolution. The design itself, components type, component placement and range
> of values, are allowed to change and “evolve” and produce new combinations
> in the design. Some of the examples in the article are of analog filter
> circuits that were not only optimized, but, new and better circuit
> topologies emerged. The same could be applied to optical design…perhaps an
> all- spherical large easily built scope design might be realized.


I have a few comments to offer.  I use optimizing routines at work and at
home.  The Algorithm of the Gods routine is somewhat slow but reliable
routine that uses a concept called simulated annealing.  The idea is to
slowly cool the working solution, allowing the solution to stabilize in the
lowest energy or optimum state.  At each step, you inject a little energy
so that if the solution is trapped in an ultimately unrewarding spot, it
can look for greener pastures.

This is very unlikely to produce meaningful results in optical design
however, though some small ideas may emerge.  Likely though these small
ieas were already thought of and simply discarded as not pertinent by lens
designers earlier.

Imagine a frothy churning ocean with pools of placid water in shaded
lagoons.  The goal is to reach a beautiful lagoon, or, in our case, a
working optical design.  We use intuition and experience to select lagoons
to study, and when successful, zoom in on a particularly clear and calm
portion of the lagoon to arrive at a nifty optical design.  The AG
(algorithm of the gods) uses the annealing process to tunnel through the
froth and seas to find these lagoons, and to eventually settle on an
optimized design.

My sense is that these lagoons are so separted and exist only in well
defined situations, and consequently, we've found them all or almost all of
them.  The AG is like a really fast moron.  It will say, gee, all
spherical, no prob, computing... wait a second... yes, I have your answer! 
And the answer will be things like one spherical surface that physically
crosses another, and other impossibilities.  These impossibilities are easy
for the experienced optician to intuit, but difficult to specify as
programming parameters.

Randomness is a difficult if not impossible concept to understand.  Most
people answer incorrectly the question "Is evolution random?".

That's not to say that nature is not infinitely more clever than we humans,
but physical laws of optics constrain optical design into well traveled
lanes, and I doubt that we will see significant innovations that are
practical coming from AG and similar application to optics.

(One way that I might apply AG is to supply it a list of common all
spherical optics, and ask it to arrange them in an optimum pattern to
achieve various goals like max color correction, coma correction for fast
newts, etc.)

 Mel Bartels

--- BBBS/NT v4.01 Flag-4