|
|
Computers > Artificial Intelligence > Genetic Programming > Research and News > pySTEP or Python Strongly Typed gEnetic Programming
pySTEP or Python Strongly Typed gEnetic Programming
Submitted 2009-09-19 03:02:45 by p33ble
Rating: 0 (0 votes) | 439 Views
Section: Research and News
Hi!
PYSTEP is a light Genetic Programming package in Python. The users can use it to easily evolve populations of trees with precise grammatical and structural constraints.
In other worlds you can set up building blocks and rules that define individuals during the evolution.
Why using rules and constraints ?
Let's take the example of the DNA of a human being which is 95% similar to the DNA of a chimp. This means that there is some strong similarity in the building blocks and the structure of the genetic code. A completely random genetic code would make no sense on a functional practical point of view because it would give some kind of unfit and disfunctional goo... I just think that, to be useful, artificial evolution has to be able to integrate basic rules that determine the shape of the individuals generated, so that we can get practical results. This does not mean we eliminate random search and we still admit poorly fit individuals. But at least we focus the search in the direction of the problem by only generating potentially useful solutions.
The strong point of this is you can really build trees that obey really complex grammar rules. e.g. you can define the order of appearance of children nodes under a parent node, their numbers, their types... etc
The package is stable and functional. I plan to add all kind of nice things to it in the next few months.
Available at: http://pystep.sourceforge.net/
Any comments highly welcomed! :) |
Discuss this submission in the forum
|
|
|