[galib] Question about TerminateUponConvergence

[Gustavo Zeferino]

Dear colleagues,

I'm trying to utilize as convergence criterion to stop, but I think I still
do not understand how to configure it.

My function is being minimized and put the following parameters for the
stopping condition:

GASteadyStateGA ga(genome);

// ... some parameters

ga.minimize();
ga.pConvergence(0.10);  // just for test
ga.nConvergence(5);
ga.flushFrequency(5);
ga.scoreFrequency(5);
ga.terminator(GAGeneticAlgorithm::TerminateUponConvergence);
ga.evolve();
ga.set(gaNscoreFilename, "bog.dat");

Wacthing the bog.dat file, it has the following statistics:

it  mean    max     min     std     div
4605 21.6706 21.731 21.5302 0.0639104 -1
4610 21.6706 21.731 21.5302 0.0639104 -1
4615 21.6706 21.731 21.5302 0.0639104 -1
4620 21.6706 21.731 21.5302 0.0639104 -1
4625 21.6706 21.731 21.5302 0.0639104 -1
4630 21.6706 21.731 21.5302 0.0639104 -1
4635 21.6706 21.731 21.5302 0.0639104 -1
4640 21.6706 21.731 21.5302 0.0639104 -1
4645 21.6706 21.731 21.5302 0.0639104 -1
4650 21.6706 21.731 21.5302 0.0639104 -1

The algorithm does not converge. I had to force a halt. Looking at the
statistics, it is obvious that the score has stagnated. Why the stop
criterion is not activated?

Att.
Gustavo Zeferino
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.mit.edu/pipermail/galib/attachments/20111123/86d01ccf/attachment.htm