NetLogo_GeneticProgramming - Programmers Heaven

#### Howdy, Stranger!

It looks like you're new here. If you want to get involved, click one of these buttons!

# NetLogo_GeneticProgramming

Posts: 1Member
[size=1]Hi all, I'm trying to figure out how does the genetic algorithm works? I don't understand what does random-seed in the program means? Is it number? A constant? If it is, why do we need it?
This is the program I'm talking about ->
http://www.cs.northwestern.edu/~fjs750/netlogo/final/gpdemo.html

• Posts: 2Member
[size=4][size=3]I understand I am replying very late and I am not quite sure if the answer to your question is still relevant to you...

the 'random seed' in this example given by you indicates a random value/path with which the turtle starts to move. In genetic algorithm, sometimes a random value is inserted to take the search process out of local minima. It may also help to speed up the search.

To my understanding, this example uses a random seed only in the first step. However, the results are clear. On one particular case with random seed 'on', I got convergence in 20 steps. The same search took 250 generations with the same population size (9).

It is simply a random path that can be anywhere on the map. The subsequent chromosomes are derived from this chromosome. The exact process depends upon the value of 'clone-chance', 'mutate-chance' and 'crossover-chance'.

On a different note, if you are trying to learn GA from basics, IMHO... it is better not to use any libraries initially. Use C/C++/Java/Perl etc. and code it from scratch.[/size][/size]
• Posts: 2Member
[size=4][size=3]I understand I am replying very late and I am not quite sure if the answer to your question is still relevant to you...

the 'random seed' in this example given by you indicates a random value/path with which the turtle starts to move. In genetic algorithm, sometimes a random value is inserted to take the search process out of local minima. It may also help to speed up the search.

To my understanding, this example uses a random seed only in the first step. However, the results are clear. On one particular case with random seed 'on', I got convergence in 20 steps. The same search took 250 generations with the same population size (9).

It is simply a random path that can be anywhere on the map. The subsequent chromosomes are derived from this chromosome. The exact process depends upon the value of 'clone-chance', 'mutate-chance' and 'crossover-chance'.

On a different note, if you are trying to learn GA from basics, IMHO... it is better not to use any libraries initially. Use C/C++/Java/Perl etc. and code it from scratch.[/size][/size]