Am 14.12.2017 um 12:54 schrieb Bald Eagle:

> I would say that might tie more into the level of intelligence idea - the slime
> mold isn't aware that it's solving a problem - it's just mechanically searching
> for food and avoiding areas where it doesn't find any.  A blind, brute force
> algorithm driven by chemotaxis.

I'm not happy with your use of the term "brute force"; that term is
pretty well-defined in software engineering, and defines an approach
where you try out /every/ possible solution until you find the one that
is correct (or, in the case of optimization problems, until you tried
all of them and can compare the "costs" of each of them to find the best
one).

That's not what the slime mold does. It doesn't try out each of the
(infinitely) many possible routes through the maze and compares their
"costs". Instead, it uses a quasi-iterative optimization approach: Start
out with one very rough approximation of the solution (which in case of
the slime mold doesn't even qualify as a route, but is just an almost
uniform blanket of slime), and modify this approximation to get ever
closer to the optimal solution.

In a nutshell, the slime mold's "algorithm" seems to be as follows:

1. Start out with a uniformly filled surface.
2. For every point on the surface, try to figure out if the point can
possibly be part of the optimal path.
3. Eliminate every point that is guaranteed to not be part of the
optimal path.
4. Rinse and repeat from 2.

Post a reply to this message