... breeding¹

Some people, including Koza [1], compare genetic algorithms to natural selection (evolution in nature) as opposed to breeding. Normally, this is a poor metaphor, because natural selection, unlike breeding and genetic algorithms, has no defined goal. Some odd variants of genetic algorithms, however, do mimic natural selection: the individuals are given the ability to reproduce themselves and are put into an environment where hopefully individuals with the desired characteristics will survive.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... programs²

Any computer program can be represented hierarchally. In fact, modern compilers translate the source code into a tree structure before operating on it, a process called parsing. The tree structures in genetic programming are often called parse trees, because, although they are not formed by parsing, they look like the parse trees used by compilers.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... one³

I want to try an initial population somewhat larger than the regular population size. This is because most of the initial population have a fitness of zero, and thus the first generation is mostly the offspring of a small number of individuals. A larger initial population could increase the genetic diversity.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.