[Push] RANDOM-SEED parameter

[Lee Spector]

No need to apologize for the questions! I think they're helpful whether they lead to changes or not.

I never use ENV instructions in evolution, and as I mentioned I haven't even implemented them in some of my versions of Push. But I think they're an interesting idea, potentially useful in some situations (e.g. for configuration), and have possible extensions (that I think Maarten Keijzer has played with) that might make them useful even during evolution. I don't have details on that handy...

 -Lee


On Jul 1, 2010, at 11:20 AM, PerPlex Ed wrote:

> I'm definitely removing the
> 
> Env.RandomSeed
> 
> instruction from my implementation of Push.
> 
> My opinion is that there should not be such an instruction. It should not be available at run time, for sure.
> I don't find useful using Push code to configure the interpreter at all, actually. But re-seeding is almost certainly a stupid thing for the great part of  Push program one can think of evolve.
> 
> I also realized that storing statistics about the generations and learning to read them is much more useful than having a "conforming" Push implementation.
> 
> That probably explains why Push language specification is often imprecise: for a real usage it doesn't actually matter. The language is even overly specified.
> 
> Sorry for bothering you with all my questions!
> 
> 
> Lee said:
>> I haven't carefully timed it to know what the actual speedup is. 
> Designing a really good test for this is probably a little tricky and 
> I've been more concerned with trying out my new research ideas than with benchmarking.
> 
> I misunderstood your "As we get into double-digit numbers of cores it makes a big difference.". I thought it had some hidden meaning :) You probably just meant that cutting the processing time to 50% or 25% is good but won't change your life while a speed up of 10, 20 or 50 times can change your routine of work and can even affect the kind of tests you may imagine to perform.
> 
> I don't think there is need of special benchmarks to measure the actual speedup. Unless you update Push to allow intercomunication between individuals or you use so much memory to store statistics, the speedup will always be proportional with the number of cores you are using, no more no less.
> 
> PerPlexEd.
> 
> 
> 
> 
> _______________________________________________
> Push mailing list
> Push at lists.hampshire.edu
> https://lists.hampshire.edu/mailman/listinfo/push

--
Lee Spector, Professor of Computer Science
School of Cognitive Science, Hampshire College
893 West Street, Amherst, MA 01002-3359
lspector at hampshire.edu, http://hampshire.edu/lspector/
Phone: 413-559-5352, Fax: 413-559-5438

Check out Genetic Programming and Evolvable Machines:
http://www.springer.com/10710 - http://gpemjournal.blogspot.com/