Re: VMs: Re: Artificial voynich - stochastic approach possible?

[sethm@xxxxxxxxxxx]

On Thu, Dec 04, 2003 at 07:13:40PM +0100, PK#01 wrote:
> > Well,  Jeff, I think you have probably now shown us enough
> > pseudo-Voynichese. I doubt whether anything more can be
> > gained from tweaking your 'tables' and procedure further.
> 
> Could Jeff try a stochastic method?
> 
> 1) The program makes a random change in his tables and produces output.
> 2) The output is checked for a statistic match with voynichese
> 3) If the random change is an improvement, then it is accepted
> 4) If the random change is a degradation, then a random choice is made
>      5) If the random choice is YES (let's say with a 10% chance) then the
> change is accepted anyway
>      6) Is the random choice is NO (with a 90% chance) the change is
> rejected
> 
> The program slowly converges to an (sub) optimal soulition and doesn't get
> stuck in small local optima. This works even for NP-complete problems like
> the travelling salesman. You might find more info under "simulated
> annealing".
> 
> I think it is worth a try. Or do you have theoretical aguments why it
> couldn't work?

A Voynichese Neural Net?

Sounds pointless, but so cool that it has to be done!  If I had any spare
time at all at the moment, I'd start working on it right now.

-Seth
______________________________________________________________________
To unsubscribe, send mail to majordomo@xxxxxxxxxxx with a body saying:
unsubscribe vms-list