Genetic algorithm and classifier system

[Wojtek Kopczuk]

Hi,
in case someone is interested:
I am currently working on implementation of classifier system
library in Swarm. It includes also implementation of genetic 
algorithm. Implementation of GA is almost done I will upload it
to Swarm ftp archive in a few days. It may be used as 
a standalone library.

GA object may be thought to be a container for collection of 
individuals.
My intention while implementing it, was to impose as few as constraints
as possible on the structure of objects it operates on. 
In particular methods to mutate, cross-over, set/get fitness are 
specified dynamically, i.e. their selectors are specified by sending
message to GA object and then hold as its attributes. GA object
doesn't care what 'mutation' and 'cross over' actually mean.

GA object may be used in a number of ways:
1. to generate new population 
2. to select a parent
3. to select an individual to die
GA has number of parameters specifying:
- generation gap - fraction of elements to be replaced when
  new population is generated
- mutation mean - intuitively expected number of bits to be mutated,
  actual value is drawn from an exponential distribution and
  sent to mutated individual as a parameter of its mutation method
- cross-over rate - how often cross-over operator is applied
Two methods used to select an individual are implemented: 
'roulette' and 'tournament'. Selecting procedure uses individual's
fitness and depends on some parameters influencing likelihood of 
being selected.
GA object may both 'maintain' its collection (i.e. remove 
and add elements to it) and perform its operations without
modifying it (for example return new collection containing
new population generated from an old one).
In addition GA object is able to display some information about its
activity and it can be used in an interactive way from GUI.

Wojtek
-- 
-----------------------------------------------------------
Wojtek Kopczuk              "There is no third way between
address@hidden       right and wrong"
IGS: wojtek                  Friedrich von Hayek 
http://www.econ.lsa.umich.edu/~wojtek
-----------------------------------------------------------