Database vs Cards (combined)

David Vaughan dvk at dvkconsult.com.au
Mon Oct 14 17:46:01 EDT 2002


On Tuesday, Oct 15, 2002, at 03:46 Australia/Sydney, Geoff Canyon wrote:

> At 8:29 PM +1000 10/14/02, David Vaughan wrote:
snip
> For 1,000 to 2,000 records, plain text or arrays will be fine. Cards 
> would be too, but I've tried creating stacks with ten thousand cards, 
> and performance suffers. With arrays or text, you can easily go to 
> 10,000 records or so, depending on what your base platform is for 
> deployment. You can stretch arrays and plain text to a few tens of 
> thousands of records with little extra effort, but eventually memory 
> becomes an issue. When that happens is your call.
> -- 
On Tuesday, Oct 15, 2002, at 01:51 Australia/Sydney, Chipp Walters 
wrote:
snip
> David, I might try to create an XML database and either write a 
> parser, or
> (if using 1.5 or greater) use the native XML stuff. If written 
> correctly the
> first time, it should be reusable for your other projects. I believe 
> the
> native parser is extremely fast.
>
> BTW -- a bit OT, do you have any more info or links to info on the 
> Genetic
> Algorithm you created recently? I saw your post and it piqued my 
> interest!
>

Geoff, Chipp

Thanks for your comments. I am comfortable with using arrays, 
structured lists and custom props so I can take that further but I will 
have a look at Chipp's XML thought before I finally go ahead. This is 
not commercial so my only pressure is to relinquish the last thing 
running in HC/Classic. I have 1.5A7 but might wait until I get my eager 
mitts on 2.0 and then decide.

Chipp, re GA

You remind me that Monte is waiting for a further response on the same 
topic (sorry Monte; I had Uni over the weekend; this may answer 
anyway). To gather information I used a Google search on some likely 
keywords and found some informative sites but did not record them. The 
approach I used was itself a selection and combination of approaches 
(how appropriate :-)):
- Relatively small population (the number of source elements up to 
twice that).
- 100% replacement after two elites.
- Best-of-three tournament selection of each mate each mating.
- Randomised crossover of genes with two children produced 
independently with replacement.
- Mutation of all factors where total mutation rate is around 0.5-1.5% 
and factor mutated is randomly sub-selected.
- Mutatory options include replacement by random new individual
- Clones exterminated and replaced.

Stopping rule is within a user-selected level of precision or when the 
user clicks the Stop button (current generation, generation of last 
change, and current values and discrepancy are updated on screen).

I found this approach robust, effective and quick enough, by which I 
mean seconds to minutes, not hours to overnight. Provided you can 
single-score the result (allowing also weighted combination of factors) 
this will work well for optimisation tasks and produces something 
end-users can employ. I had earlier considered the IP aspects but I am 
pretty well occupied by my day job so I am happy to send a copy of my 
sample stack as acknowledgeware (if you use my ideas, mention it) and 
to provide some assistance if desired.

regards
David
>
> regards,
>
> Geoff Canyon
> gcanyon at inspiredlogic.com
>
> _______________________________________________
> use-revolution mailing list
> use-revolution at lists.runrev.com
> http://lists.runrev.com/mailman/listinfo/use-revolution
>




More information about the Use-livecode mailing list