Genetic Algorithms
- Genetic algorithms (GAs) are another class of algorithms that
can learn.
- The idea is that you describe solutions to a problem as
a series of strings of elements.
- Each solution is evaluated and the best solutions
reproduce.
- The reproduce by taking parts of their string and recombining
them with parts of other successful strings.
- You can also modify strings by simply changing elements of the
string (mutation).
- There's some argument about the name, but the idea is
pretty similar to sexual reproduction.
- Roughly, you search for a good solution based on randomly generated
poor solutions that are recombined to preserve the best bits.
- One problem is coding the solution as a string of elements.
