why do we crunch collatz?

[MindCrime]

Can someone explain to me what is so amazing about the question that collatz is asking? It's a pretty simple concept in my mind, I guess the first way I "boxed it up" was HoTPO in its simplest cycle is 4,2,1 and asking the same questions from any higher integers is purely an offset, the integer I start with is my coefficient and the instructions are built in to factor it out. We also have a bit of logic in it, if even then divide, if odd then multiply and add 1, the answer is in the question

HALF or Triple Plus 1 2=half 4=triple+1 it's just factored to smallest terms, yeah it climbs up and down before it's all factored out but that's the nature of climbing on odds and falling on evens but there's more evens because the equation asks for evens thus it eventually falls all the way to simplest terms 4,2,1

is there something greater to this numerical oddity?

is it because it's easy credit?

[cineon_lut]

Sadly, yes. It's easy, good credit. If we hope to maintain a good team overall score, it needs to be on our list. The project from a technical standpoint is great: The apps are well written and efficient, easy to run, generally reliable and it always has work. Pay is good too.

Like most math projects though, it's scientifically bankrupt.

Disregard everything I said if one day someone on BOINC proves the conjecture wrong. :)

[DrPop]

Most things in BOINC are a little strange. Like . . . why do we crunch Collatz? :) Well, for starters, it's on par with several other projects - as in not really much scientific value coming out of it, but more of a "proof of concept" I think. Secondly, the project is VERY well run, Slicker (one of our team mates who runs the project) responds to changes and questions near immediately, and is always on the ball with new and improved GPU apps.
Third, in the grand-scheme of BOINC stats, it is a fairly large project. Not the biggest, and not the highest credit, but it's up there in the top 1/3 for sure, and we ignore it at our own peril if we want to maintain a competitive score.
Hope that helps a little! :)

[c303a]

Remember that you as an individual do not have to run any project you do not want to. The team as a whole, is a crunch what you want team. I usually just run the medical science projects, but I will sometimes run different projects to help the team during a challenge. If you want to run a project, just do it, if not, don't. Pick the projects you like and give them h***.

[myshortpencil]

Can someone explain to me what is so amazing about the question that collatz is asking? . . .

is there something greater to this numerical oddity?

is it because it's easy credit?

I asked the same question on this forum a while ago, and the best answer I got is that Slicker -- a team member -- runs the project and his endeavors to write ever better and more efficient codes for a wide variety of CPUs and GPUs ends up helping other projects. Think of Collatz not just as math, but as a testing ground for innovations in distributive computing. To what extent does the expertise gained in Collatz help other projects? I don't know, but you could send a private message to Slicker, or wait a while to see if he replies here, to find out. :)

[Mumps]

Well, keep in mind as well that Collatz was (I believe) the first project to implement *any* code to support ATI cards. And as such brought a lot of FLOPs to the BOINC arena in general.

Also, while some may disparage many of the "pure math" projects as scientifically bankrupt, many of them are hunts to prove (or disprove) many hypothesis that are the foundations of many things important in modern life. If some of these conjectures prove to be invalid, it can theoretically dump modern science on it's ear. Just like the expected discovery of the effects of gravitational waves (one of the basic sub-projects at Einstein@Home) helps to solidify Einsteins theory of relativity, many of the numeric conjectures help solidify our understanding of the universe around us. In the case of Collatz, it's one of a set of math tenets (http://en.wikipedia.org/wiki/List_of_unsolved_problems_in_mathematics ) that are theoretical truths, but no proof exists.

[MindCrime]

These are all really good answers, and I appreciate all the feed back. I personally don't believe it will be proven wrong as every integer is a factor of 1,2 or 4 and the way I see it is it's one of those things that look really cool when graphed but on paper seems mundane. A couple of responses were about the program and how it's a 'test bed" for apps! This reason right here is good enough for me to run it, and my computer seems to run it well compared to other projects. I had the resources turned down on it but I'll give it an even share again. I also didn't have both gpus running when I started thinking about what my project priorities were, now I do and Im eager to compare the intel and nvidia gpus.

[Slicker]

Collatz has no purpose other than pure math and for my entertainment. (I really don't find math all that fun, but don't tell those who crunch the project solely because it is "for the math"). On the other hand, I do try to be the first to implement various features such as ATI support, Intel GPU on OS X, OpenCL for CPU, multiple-size WUs matched to the hardware, etc. But, it is a hobby for me and is totally funded by me. None of your tax dollars via government grants are being wasted searching for little green men. IMO, looking for pulsars bazillions of light years away that will never impact us here on Earth is of no more use than solving a math problem with no real world application either. Collatz searches for a single number that will prove the guy who dreamed up the conjecture 80 years ago wrong. Crunch it if you like. Crunch other projects if you like. I do both myself.

Hopefully, when other projects are down, Collatz is still be up. When other projects can't generate enough GPU WUs for the users, Collatz can. When other projects decide that the SETI ratio of GPU to CPU speed is the measurement standard and enable creditNew, Collatz will still be awarding more credit for more work and less credit for less work. So, if you underclock your GPU, you will get less credit than someone who doesn't. If you overclock your CPU, you will get more credit than someone who doesn't. If your 64-bit OS does twice as much work in the same amount of time, it will get double the credit. It will not be "normalized" as it does under creditNew. So, in some ways, it is a stand "against the man"!

Note: There is a development group working on fixing various shortcomings of creditNew and I gave them a long list of items on which to work. They are NOT the standard BOINC development team, so whether DA allows the changes to be merged into the master source code when they finish will be anyone's guess since DA never got anyone's input when he designed it or implemented creditNew. I could go on a real rant about creditNew at this point, but this post is already too long for todays' ADHD youth. ;-)

[cineon_lut]

I should now start another thread "Why DON'T we crunch collatz?"

I'm guilty of losing some GPU power over the last month, but hoping to get back in business soon. HOWEVER we're getting killed on this project. L'AF is smoking everyone on this project now according to FDC, and their Collatz output has them neck and neck with us on daily combined output. So roll call: where are YOU pointing your GPUs?

I'm switching from GPU Grid back to Collatz for a while.

[Mumps]

I have one GPU that can't run any other GPU project than Collatz. So that's where it lives. Other than that, I'm GPU poor...

[cineon_lut]

I have one GPU that can't run any other GPU project than Collatz. So that's where it lives. Other than that, I'm GPU poor...

I'll trade ya... ;)

[rgathright]

MY OPINION AFTER READING THIS THREAD:
1) You don't have to crunch Collatz and it has little scientific value but it is very stable because it is run by an active administrator.
2) I crunch Asteroids@home for its "Space/Astronomy" value but with someone as capable as Slicker would administer it!

CONCLUSION: From a 10+ year BOINC/SETI participant, Slicker is the man! :cool:\:D/\:D/:)>-

[finalfugue]

the way I see it is it's one of those things that look really cool when graphed but on paper seems mundane.

G G G Eb F F F D

Boring, right? Mundane. Seemingly pointless. Random. In this context, yes. When graphed in another context, not so much (https://www.youtube.com/watch?v=rRgXUFnfKIY).

Why do I crunch Collatz? Because it's a chance to study an interesting pattern! It's not my project, time, or hardware, but I dislike the idea of saying we're trying to "disprove" the conjecture. It's not a football game. There need not be a winner and a loser. While our computers are fantastic at completing large amounts of computations, and in recent decades have even become passable at pattern recognition, they are still (thankfully for our survival) pitiful at discovering a USE for the patterns and knowledge they unearth. Our computers complete this simple arithmetic task over and over again because Slicker's program tells them to. They feel no tug of their governing idée fixe.

We may yet disprove the conjecture. Though we think we see a pattern that extends to infinity, there may be a pattern over the pattern, the first member of which is an enormous figure. Or it could just be one odd alignment that throws a wrench into the whole thing. Then what do we do? Where's your integer god now, eh? Even if we never disprove the conjecture, the study of it may prove useful to just the right human with just the right pattern of neurons. We're making this huge record of figures we know eventually reduce to 1. Some do it very quickly. Others take a little longer. Others still take a MUCH longer time to reduce. The study of this record could someday undermine a queueing architecture for financial transactions, a memory accessing algorithm, or a 15 Grammy-winning dubstep album. I don't know. I don't see anything in it. The right person might.

Suppose that person then uses our little dataset as the underpinnings of a new digital architecture. That person then goes on to win the Nobel prize and all future computers manage their enormous power based on this new algorithm- including those at the Department of Defense. Years go by and humanity marches on fighting, loving, crunching... until the first computer hits a figure at 10^25839427864. It tries to queue something but it never returns, or is so many steps away it "practically" disappears to the rest of the system. The information, no matter how trivial or small, falls into an endless computation sequence. Other computers begin to hit the same figure. An automated NORAD monitor registers this as a sign the stations have been eliminated. What computers are still functioning complete their retaliatory launch sequences before becoming useless bricks (or CreditNew servers, same difference).

THE WORLD ENDS BECAUSE YOU DIDN'T SEE WHY IT WAS IMPORTANT TO "ENDLESSLY" COMPUTE COLLATZ! (https://www.youtube.com/watch?v=n4vz5iSVoYw)

Do you get it now?! Does not being consumed in a nuclear holocaust make "sense" to you?! Good.

Oh, yeah, and also Slicker's the man. The project is well-run, has an active community, and blazes a trail for new hardware/driver/software uses and implementations. And it keeps my Intel iGPUs from killing themselves out of boredom.

In other news- I'm bored at work.

[MindCrime]

There is a pattern... it's a game and it's rigged. Triple any odd number and you will get an odd number and since we started with an odd number we need to add 1. Now we have an even number, and we should divide by 2. You will divide by 2 far more times than you will 3x+1 because 3x+1 always gives you an even number and every other even number will divide in half by an odd number. So your outcomes given to you by the conjecture are 75% even numbers and if you divide by 2 75% of the time and 3x+1 25% of the time you're always going to get to 1.

I think of it as attach 1 of 2 coefficients to any integer, 1 if its odd, 2 if its even. As long as it works with with 1 it'll work with anything integer because they're all divisible by 1. Let us use 1; 1 is odd so 3(1)+1=4; 4 is even so 4/2=2; 2 is even so 2/1=1 this is the most factored it can get, anything else we start with is just a multiple.

That said I too have it as my intel gpu filler app.

[finalfugue]

Of course that's how it works... but good luck getting George Clooney/Harrison Ford to star in *your* story. I'm off to build my Collatz Doomsday machine.

It will require double-precision and a custom closed-loop cooler.

Sent from my HTC6500LVW using Tapatalk