[galib] Odd thing happens to GALib

[Liekens, A.M.L.]

Genomes that have been copied (without interference of mutations or recombinations) are not re-evaluated (since the fitness of individual is assumed to be the same generation after generation). In the case of a static fitness function, this is just fine. If you want to make a fitness function that is dependent on other individuals in the population (game theory) or generally dependent on time, this could be annoying, but you don't seem to have such problems.

Anthony,-

-----Original Message-----
From:	galib-bounces at mit.edu on behalf of Xiaowei Tan
Sent:	Tue 5/17/05 05:55
To:	galib at mit.edu
Cc:	
Subject:	[galib] Odd thing happens to GALib
I am using GA Lib for my project. I just found something odd with GALib:

For a genetic algorithm defined with population of N and generation of M,
the total number of objective function evaluation should be Nx(M+1)
(Considering ga.initialize(), where GaLib generates the first generation).
However I found the number of data points is always less than N(M+1). After
I put a global variable to trace the number of objective function
evaluations, I found GA Lib starts off with evaluating every individual in a
population, start to skip more and more evaluations after a few
generations...

I am perplexed. Can someone here explain why?

 

Thanks!