Constant n-Search for k*2^n-1

[Cruelty]

This is a rhetorical question. Does it have anything to do with my previous statement?

[em99010pepe]

I see people here willing to donate some CPU power and primes to the RPS project and really I don't understand Kosmaj position.
If only Kosmaj decides what primes should or shouldn't go to RPS I prefer to release all my reserved ranges. I'm in standby...

Kosmaj, you need to be more open minded and less aggressive on your statements.

Carlos

[Cruelty]

AFAIK, this relates to some issues on the level of TOP-5000 prime list, however sometimes Kosmaj is indeed a little bit harsh with his opinions

[gd_barnes]

Originally Posted by Cruelty

Given the lack of definition of the word "many" for no 1, I guess we can report just any riesel prime as part of RPS

That's what I've been doing. I just figured that it was a nice gesture to the project to report my top-10 twin and now two top-5 quads to the project even though they were fixed-n searches. After all, it was RPS that got me excited about prime-searching again!

I also do not understand Kosmaj's position on this. But I've had it out with him before and nothing productive comes from it so I've now just chosen to focus on prime-searching for the greater good. If he doesn't want primes reported to RPS that don't fit specifically the criteria of the project, then that's what I'll do.

Carlos, we certainly wouldn't want to lose your new awesome CPU power but I could understand if you took it elsewhere.


Gary

[SB2]

Originally Posted by gd_barnes

If he doesn't want primes reported to RPS that don't fit specifically the criteria of the project, then that's what I'll do.

Gary

Part of the answer to this is contained within the email you recieved from Prof. Caldwell (in a round about manner). When the projects are set up on his Top 5000 website he askes that the project only submit primes within the defined scope of the project.

[gd_barnes]

Originally Posted by SB2

Part of the answer to this is contained within the email you recieved from Prof. Caldwell (in a round about manner). When the projects are set up on his Top 5000 website he askes that the project only submit primes within the defined scope of the project.

I never received anything from Prof. Caldwell about project criteria so I don't know what you're talking about. Prof. Caldwell was concerned on one of my quads that I had reported the +5 and +7 primes under LLR instead of Primo because using LLR, they could only be probable primes. He didn't care that I reported it under RPS.

Regardless, even though I didn't get anything from him voicing a concern about the project that I reported the twins or quads under, I hear what you're saying about him only wanting them reported under the scope of the project.

The problem is that 2 of the 3 criteria of RPS are so broad and ambiguous that it's not easy to tell what should be RPS and what should not. The question is, how do we specifically define "k's that can produce many primes in the given range of exponent n" and "low-weight k's that produce a very small number of primes"? Is the first for weight > 2500 or 3000 or 4000? Is the second for weight < 1000 or 500 or 250? If not, than what is the criteria?

If you use the above criteria, then my twin prime k (weight 2883) and one k out of my two quads (weight 2838) fit the first definition and so I was correct in reporting them under RPS even though they were a fixed-n search. Both are well above the average Nash weight.

The way that this project is different than almost all of the others that I've seen is that its scope is not clear. Also, it kind of 'pseudo-expanded' itself when Karsten began collecting data from all available sources for Riesel primes. After all, there's no doubt what to report for the GIMPS or Rieselsieve or SOB or 12121 or many other projects. But there's a big doubt for RPS and that is where the problem comes in.

I think it should be up to the creators of the RPS project to come up with a more specific definition of it so that we can finally end this problem. Kosmaj, Larry, Amphoria, anyone else...any thoughts?


Gary

[SB2]

Originally Posted by gd_barnes

I never received anything from Prof. Caldwell about project criteria so I don't know what you're talking about. Prof. Caldwell was concerned on one of my quads that I had reported the +5 and +7 primes under LLR instead of Primo because using LLR, they could only be probable primes. He didn't care that I reported it under RPS.

Regardless, even though I didn't get anything from him voicing a concern about the project that I reported the twins or quads under, I hear what you're saying about him only wanting them reported under the scope of the project.

Gary

True, my explaination was a bit off the mark, I was trying to imply that if the projects reported primes started wondering off in to areas not covered by the projects defined goals/search the project admins would be getting an email.

RPS has added to the projects "scope" when the "low-weight" k's were added and again when the "small k's" were added. So expanding the project is not unheard of, to satisfy the interest of members wishing to venture into other areas. Not advocating any expansion ... just saying.

[Kosmaj]

RPS has always been about fixed k search (of k*2^n-1 primes). If the word "fixed" is missing from the scope declaration I'll added it. Anyway the scope declaration was not intended to withstand scrutiny by lawyers

About the subprojects:
1) k's that can produce many primes: We used to try many k's and test them to n=10k or 20k and 50k and if there were many primes found there, we expanded the search to higher exponents. Nash weight is not always the best indicator, it's better if the number of primes up to a certain n is known (i.e., tested for primes). Many such k's were found initially by Shane (15k) and later by Larry and they are listed on the Choose your own k page (we stopped working on those found by Shane to avoid a potential duplication of efforts).

2) Low weight: Initially championed by Thomas IIRC, here we can maybe set a numerical threshold on the Nash weight, the idea is that by LLR-ing a k with a small number of candidates left after sieving, one can reach high exponents quickly and, if lucky, find a large prime.

3) k<300 is self-explanatory, but k's already reserved by other projects and individuals are excluded, and new reservations by people not associated with RPS are welcome.

As SB2 noted, with respect to the "15k" project we did a number of extensions already: allowing k's not divisible by 3 and 5 in (1), and adding (2) and (3). I don't see any reason to extend it any further considering the fact that we already found 821 primes (on any list) and 455 on the current Top-500 list. If we extend further I think it will be nice to consider some "exotic" primes like Mersenne norms suggested by Cruelty (we already agreed to include them but no new primes were found).

And to reply to Gary: when Karsten added all k's up to 10,000 to the stats pages he didn't expanded the RPS projects. He was just creating new stats pages, but not reporting any new primes. Anybody is free to create whatever stats/web pages he wants within or without a specific project.

And to try to reply to Carlos: Top-5000 "projects" are not "teams". If they were teams we could report any primes including k*2^n+1, and all other forms, but they are not.

BTW, I'm a nice guy (I think) but every project needs some guidelines and somebody has to remined the others about them.

[VBCurtis]

Some part of the problem here is that we chose a rather broad name for this new project when we split off of the original 15k search. If we had chosen something more descriptive of the plan at the time, we'd be something along the lines of "special-fixed-k search for riesel primes", with special defined in the project's scope.

When newer people see "riesel prime search" and an RPS forum, the natural assumption is to use it to coordinate and report *all* riesel prime work. The forum is indeed the best place for this coordination, and the 15k.org website the best place to manage and present the data, but the RPS project code on top-5000 is meant for the specific project as originally set out.

I, like Carlos and Gary, like to see RPS as a team sort of thing, but top-5000 does not view projects in that light.

Finally, my weightless opinions on fixed-n: the n, k-space is 2-dimensional. We have chosen to organize work by single k's and ranges of n. Reversing that organization means some duplication of work where a reserved n necessarily overlaps some k,n pairs that have already been searched, or will be searched via fixed-k search. As such, despite the sieve-rate boost, I think it wasteful to organize or reserve fixed-n searches. The increase in sieve speed is usually offset by the increased LLR timings for large k in the sieve ranges; moreso when one decides to "take advantage" of the fact that the sieve speed is hardly affected by size of k-range. The searcher ends up with LLR times ~doubled for a size of n compared to k<512, while the sieve maybe get a mere 3 bits (8-fold) deeper (resulting in something like 6% fewer tests). Finally, there's the chance someone doing a fixed-n search snags a prime from a large k already reserved- whoops!

In short, fixed-n searches for riesel primes are likely a bad idea. For twins/more esoteric types, they're necessary, but for straight riesel searches, the speed gains are illusory.
-Curtis

[VBCurtis]

Perhaps we can ask Prof Caldwell for a new project, the "riesel prime forum searches", where our stated goal is to find, organize, and report riesel primes of any sort not already outlined in other projects. This would include k>1001 (since k< 1002 is primesearch or RPS), fixed-n searches, twins/triples/quads/Sophie-Germain/etc like Gary found.
Prof Caldwell may reject such a definition as too broad, but this idea illuminates the difference between the purpose this forum serves, and the RPS project's definition vis-a-vis the top 5000 page.
This also sheds light on why gary and kosmaj do not see eye-to-eye on the scope of our project. www.15k.org and the forum are *not* the RPS project; the latter is a subset of the former two. Kosmaj, I think, does not oppose data gathering and organization for any riesels, but the RPS top-5000 code is what he has so staunchly defended from expansion of scope. Gary's interest is in the data management and completeness, while Kosmaj's more in the top-5000 caretaker of the project sense. In this sense, both are quite right in what they want!
I wish this distinction had been made months ago; it would have saved much discord in the forum. Am I right?
-Curtis

[em99010pepe]

In some way RPS is a team because it is ranked.
Here's my humble opinion: acting like a team there's competition between members, members will start teasing each other (good thing) and as consequence you will buy more machines. Who gains? The RPS of course.

I really don't have fun hunting for some primes if I don't see RPS as a team....

BTW, You think you're a nice guy . Seriously, you're a good leader, you won. I'm back to LLRing.

Carlos