Are the recent advancements for GIMPS (VDF, Gerbicz check, Jacobi check) worth publishing?

[M344587487]

Originally Posted by pinhodecarlos

Chris, moderate your language. Publishing a paper is a deal and final owner is the journal.

I bet you a quid that I'm more offended by the offense you've taken at an inoffensive comment, than you are at the comment you're offended by. That was incredibly moderate language given the racketeering cesspool being described, how much more moderate could it be?


Not that this is a vote but I think the whole system can shove it, moderately speaking. Codify the findings into the normal publishing format, self peer review and self publish.

[Batalov]

Originally Posted by chalsall

Separately (and very importantly), are others comfortable with this idea?

No

[ixfd64]

Originally Posted by henryzz

A peer review for publication in a journal would add a lot of confidence to our results.

I agree. As far as I'm aware, the Gerbicz error check is only discussed on this forum. Academics generally don't consider forums to be reliable sources.

[Prime95]

One aside. The EFF requirement for claiming the 100Mdigit prime award is publication in a peer reviewed journal and I do not enjoy writing.

The 1M award was fairly easy - a review of how GIMPS runs an efficient search using Internet volunteers for the horsepower.

When I had to do another article a half dozen years later, I had to fill several pages based solely on the premise of "GIMPS found another one". Without a connection to a Journal editor we would have had difficulty getting that published.

That argues for waiting, so that there is some "meat" to put into a 100Mdigit journal article. On the other hand, the 100M discovery is not likely to happen anytime soon.

[preda]

Other candidates for inclusion:

1. Robert presents a great simple trick that has the potential to reduce the cost of P-1 first-stage by a factor of around 10x when ran in conjunction with PRP (improving from my trivial idea offering a 2x gain):
https://mersenneforum.org/showpost.p...6&postcount=14
(I'm currently working on implementing this)

2. PRP-1,
an original idea that came to a dead-end. Dead-end or not, in time I did gain an improved understanding of its behavior and the trade-offs involved, which may be interesting to share.

[preda]

Originally Posted by Prime95

That argues for waiting, so that there is some "meat" to put into a 100Mdigit journal article. On the other hand, the 100M discovery is not likely to happen anytime soon.

The 100M digit prime.. seems so far as to be almost out of reach.. 20years? 50years?

But the very next Mp, that's within reach. Not as glamourous as the EFF prize, but feasible.

Anyway, publishing may help with awareness of the project.

[kriesel]

Originally Posted by chalsall

A sincere question: Would you be willing to lead on the paper?

In the Social Sciences, all the authors are named alphabetically ...

In this case, clearly others did the "heavy lifting". You would simply be documenting their work.

Thanks for the implied confidence in my qualifications. And no.
I would be happy to contribute in any way that forwards the action, subject to time constraints and other priorities, and being wary of reliance on me for the deeper math areas. (I've done spell and grammar check etc. for others on occasion. Fft multiplication gives me a headache, and doing it in IBDWT seems like magic. The PRP proof related material is like chinese food; understanding is gone in an hour. The new P-1 thread is after a single skimming, like a jet plane or perhaps hypersonic missile flying over my head. I suspect I have some company in those.)

My experience with peer reviewed papers is as a contributor in the work, not lead author or primary contributor. In those, author listing order was by largest contribution first and in descending order following. That order is generally settled by discussion among the coauthors. These papers were in engineering, not mathematics. (So my Erdos # is probably infinity.)

Depending on the timing of Mlucas implementation of the VDF etc., Ernst might be in the mix too. His writeup of the fft method of multiplication seemed to me a fine piece of writing.

Let's not forget
forum username error (don't know real name) identification of the Jacobi symbol check for LL sequences
Pavel Atnashev (VDF/ PRP proof)

This whole discussion has set me thinking of writing a longish blog post about the history/evolution of GIMPS software. If any part of it was useful in preparing an article or two for publication, so much the better.
I've been wondering if bundling the recent progress into one paper is the right way to go. Perhaps the math basis in one, and the software engineering in another, referencing the math paper. Plus releasing one considerably after the other and in a different journal/field increases GIMPS exposure.
As to waiting for a 100Mdigit prime's discovery, if the crude projection of https://www.mersenneforum.org/showpo...5&postcount=11 is anywhere close, some of us will be unavailable then.
We should write our history up as it happens while people are still available and clear headed to ensure accuracy.

A hypothetical great outcome of a publication would be reaching some mathematician who sees the project's approach summarized, and identifies, say, some new method of factoring that reduces the number of needed primality tests.

[preda]

Originally Posted by preda

Other candidates for inclusion:

And George's design of super-precision fast sin/cos primitives in the FFT context may be of interest, with help from Ernst who contributed the initial optimizer based on Chebyshev polinomials IIRC. IMO such work is worthy to be integrated in math libraries (such as the ones offered by AMD and I imagine by Nvidia for GPUs), and proposes a new API for trigonometrics that allows to extract the increased precision.

[kriesel]

Originally Posted by henryzz

Define low Erdos number.

The minimum possible Erdos number will increase over time, as succeeding generations of mathematicians die off. Paul Erdos is currently unavailable for new publications, having died the year GIMPS was created.

[Uncwilly]

Originally Posted by kriesel

The minimum possible Erdos number will increase over time, as succeeding generations of mathematicians die off. Paul Erdos is currently unavailable for new publications, having died the year GIMPS was created.

Thus 599

[chalsall]

Originally Posted by kriesel

This whole discussion has set me thinking of writing a longish blog post about the history/evolution of GIMPS software. If any part of it was useful in preparing an article or two for publication, so much the better.

I think that would be quite valuable.

It would be a peer-reviewed (and, technically, published (although not "Published")) collection of all the very interesting things various people do here. And, importantly, when each step happened.

I often point to this Forum as a great example of what online collaboration can do.