Hi Peter,
Sorry for not getting back to you sooner. I've just downloaded the new code and put it
through some tests and it seems to be working fine.
The only slight problem is that on line 116, or thereabouts, there is a print command which
outputs the random number generated within 'init' regardless of the verbosity options.
Cheers,
Kevin
On Sat Sep 10 04:21:49 2011, peter.jaquiery@ihug.co.nz wrote:
Show quoted text> Hi Kevin,
>
> I've just uploaded version 1.004 of AI::ParticleSwarmOptimization
> (note the
> name change) containing your plateau early exit code. Note that I've
> made
> the mean calculation faster and less memory hungry by using a circular
> queue.
>
> If you have the chance to check the changes and see that it works as
> you
> expected I'd appreciate it.I've attached the module in case you get a
> chance
> to look at it before the new version turns up on CPAN.
>
> If you are happy with the code let me know and I'll close the bug.
>
> Cheers,
> Peter Jaquiery
>
> ----- Original Message -----
> From: "Kevin Balbi via RT" <bug-AI-PSO-OO@rt.cpan.org>
> To: <undisclosed-recipients:>
> Sent: Thursday, September 08, 2011 3:30 AM
> Subject: [rt.cpan.org #70790] Detection of 'best' Best-Fit before the
> optimisation completes
>
>
> > Wed Sep 07 11:30:06 2011: Request 70790 was acted upon.
> > Transaction: Ticket created by KBALBI
> > Queue: AI-PSO-OO
> > Subject: Detection of 'best' Best-Fit before the optimisation
> > completes
> > Broken in: (no value)
> > Severity: (no value)
> > Owner: Nobody
> > Requestors: KBALBI@cpan.org
> > Status: patched
> > Ticket <URL:
https://rt.cpan.org/Ticket/Display.html?id=70790 >
> >
> >
> > I needed the ability to detect if the 'optimum' had been reached
> before
> > the number of iterations
> > had completed.
> >
> > I have attached a version of the module in which I've implemented
> this
> > option (disabled by
> > default for backwards compatibility) which detects if the fitness
> value is
> > unchanging of a
> > specified number of iterations. Also I built in the ability to
> adjust the
> > sensitivity, through
> > number of decimal places (although arguably this should be
> implemented as
> > significant figures)
> > as well as allowing the opportunity to set a 'burn in' time before
> which
> > the optimisation doesn't
> > check for 'plateaus'.
> >
> >
> >
