This site is supported by donations to The OEIS Foundation.

User talk:Michel Marcus

From OeisWiki

Jump to: navigation, search



While writing a small script for your sequence A169974, I ran into a small problem.
Definition says it is: Sum_{i=0..n} { 2^C(n,i) } + 1 - n.
Instead of getting 2, 3, 6, 17, 96, 2111, 1114238, 68723671293, ...
I get 3, 4, 7, 18, 97, 2112, 1114239, 68723671294, ...
as if the formula was Sum_{i=0..n} { 2^C(n,i) } - n.

Thanks for the feedback. It looks more like f(n)-1, but it'll take me a little while to remind myself of enough context to make sense of what I wrote back then. I'll try to take a look this weekend. Hugo van der Sanden 03:37, 24 August 2013 (UTC)

Records are scant: it doesn't look like I ever wrote code for this, the only record I can find is my email to seqfan proposing it (after which Neil added it), but there was no followup on the list.

In any case, the intent was to count "abstract simplicial complexes over n vertices for which all facets have the same dimension" (which I originally gave as the sequence name), so I think your proposed correction to the formula is the correct fix (and I assume the actual name and the Pari code will get matching corrections). I have verified that the corrected formula gives all the listed values. Hugo van der Sanden 16:37, 24 August 2013 (UTC)


You have suggested for publication, I believe, A228850. Is your notion that this of the three (The title one here is its companion and there is a 'main' sequence--endprimes of a record and the record) should be approved before the others are considered? Is there something that needs to be done with A228851 other than to wait? All it does is list terms and point to A228850 (and certain other minimally essential things), but I have no idea if this is acceptable.James G. Merickel 16:40, 24 September 2013 (UTC)

I have reviewed A228850 because I thought I understood it and also since I was able to write a script that gave same data.

I understand that A228851 are the concluding primes of the strings of primes. On the other hand I don't really understand why A229056 is including ties, because for me this brings a discrepancy between (A228850 and A228851) and A229056.

I don't know why A228850 is not going further.

It's not including them other than in the COMMENTS section (and that the PROGRAM will also give them). I changed all 3 sequences at the same time. This maintains the completeness of my research without making it part of the sequence proper. As far as '...going further' is concerned, it maybe worth commenting upon, but I am not ready to say very much on it.James G. Merickel 11:49, 25 September 2013 (UTC)

Oh, I see you have your own program there (Mine is at A229056). This changes what I think you might mean. Are you saying you have your own run going and that you don't know why you don't have more terms to add yet? Over the long run for very very large values, the differences SHOULD look uncorrelated, with a mapping of the primes to {-1, 0, 1} as a decrease/equality/increase in differences going to IIDs with probabilities P(-1)=P(1)=.5 and P(0)=0. So 52 or 53 will only occur about every two or four quadrillion primes in the large. The next record may not be computable at all. If your search is faster than mine--I haven't yet even added an order of magnitude to the search--and you intend a continued run, I could first of all use where you are in search in adding to COMMENTS and second of all shut down my search. So, please let me know.James G. Merickel 12:04, 25 September 2013 (UTC)That is, the larger the primes we are looking at the more they will look like flips of a fair coin, and we are talking about a huge number of alternating flips of a fair coin. So, the longer the wait the longer it is expected to be, and ultimately with too long a wait.James G. Merickel 15:37, 26 September 2013 (UTC)

No I'm not runnin the program. Was just satisfied to get a few terms matching yours. --Michel Marcus 21:03, 26 September 2013 (UTC)

Okay. Well, I have already added something to the COMMENTS of A229056 about the search limit and the 'more' keyword. Nothing very formal. Just saying that a new term is not all that is wanted, in case it is too hard to get.James G. Merickel 12:30, 27 September 2013 (UTC)

Any idea why these aren't published yet (and nobody is saying anything there about recommended improvements)?James G. Merickel 11:32, 3 October 2013 (UTC)

No, no idea. Sorry. --Michel Marcus 11:50, 3 October 2013 (UTC)

Since I am here below, I am more about the task right now of improving old sequences of mine than of adding new or improving other people's, so after a time sometime soon I will have more on this (I expect a better search limit only, but who knows), which did get approved eventually with little change to what I had at the time I was inquiring. In case you or anybody else looking here has an interest. (And it is much appreciated for whatever extent your own input may have helped at the time.)James G. Merickel 07:56, 28 July 2015 (UTC)


tried to get your Pari script to work but no luck, sure it's something simple. perhaps I can email you my script? It is just producing 0's for me ...--Bill McEachen 14:07, 3 November 2013 (UTC)

Sorry Bill, I haven't looked at this page for some time. And I did not receive a notification.
Maybe I can check something somewhere so that I get a mail telling me you've written something for me. Do you know where ?
Or you can email me through "email this user", my mail account is nearly always open.
BTB, yes, please send me your script, I"ll have a look.
Best Michel.
well, I see CRG has posted an alternate script that is grossly more compact than mine and that works fine.--Bill McEachen 15:03, 2 December 2013 (UTC)


(replying here as I cannot log into at the moment). I can post the data from home, but I was merely commenting that the numerical ln(a(2)) through ln(a(8)) were interesting. They "seemed" to follow a pattern, which broke at the last term I gave. I have no idea why those (7) terms would exhibit such numerics particularly, though I only an amateur and maybe miss something basic...--Bill McEachen 19:41, 8 April 2014 (UTC)

A245281 / A244801

Hi Michel,

I sent you an E-Mail to the address attached to your OEIS Wiki account.


Felix Fröhlich 12:39, 13 August 2014 (UTC)


Hi, Michel.

Please could you take a look at the sequence A257077 (draft)?

I don't know if would better to change the description as Danny Rorabaugh has suggested (it has seemed good to me) and if there is anything else to improve.




Thanks Marcus, you are correct. Keyword "fini" should be added to A153753. Reviewed the keywords so I can apply all of the correct ones next time.

Doug Bell 04:43, 6 June 2015 (UTC)


At a sequence I am editing, I made a remark about adding terms to the c-ref sequence recommending that 'hard' there be changed to 'more' if only a single term (one I provided) was added; and you just changed the one I am on to 'more' where I think it nearly inconceivable. I wonder if you mistook my comment on the other sequence (which no longer is hard to add terms to, having only been edited when submitted over 10 years ago). Having 'more' at mine probably harms nothing (but I doubt it will affect anything, unless you mean for it to be 'more' in advance of hardware improvements, i.e. that it's a question that should be answered eventually whether or not it is really much possible now).James G. Merickel 06:47, 28 July 2015 (UTC)

I occasionally see 'more' used where a next term is nearly impossible to get soon, but I usually think of it as an encouragement to try to. My question about whether I may have had a different faster program at one time because it seems I would not be able to generate the current last term shouldn't be taken to mean a best programming effort is apt to go further. I do recall that however I got that far it took a very long run (but wonder if what I have as a program isn't more like years to run).James G. Merickel 06:59, 28 July 2015 (UTC) FWIW by way of explanation, a program I've written may disappear into a kind of blackbox while running, aside from it being in my head ideationally for some short time; and if I merely intend to record it after a run (it being essentially unavailable during), I have lost much to crashes over the years. So at the sequence in question I may merely have pumped out something mediocre in terms of efficiency as a substitute (I also think wanting special side data after the fact may have played a part).James G. Merickel 07:19, 28 July 2015 (UTC) The program doesn't look so slow that it wasn't responsible in current form, after all. Even if there is a time doubling it looks like a few months would get to the final known term. Especially if I take the current load into consideration. It's about a 1000-to-2000-hour task.James G. Merickel 07:35, 28 July 2015 (UTC)

OK --Michel Marcus 07:37, 28 July 2015 (UTC)


Michel, you did a correction of this back in 2013. The sequence title is "sum-of-digits is also a perfect power." The original sequence were perfect powers whose digit sums were also perfect powers. After editing it is now any integer whose digit sum is a perfect power. I believe either the title or the sequence needs to be changed. IMO the "perfect powers whose sum of digits is also a perfect power" sequence is more interesting and ought to have a sequence (whether it is this one or a new one). Dana Jacobsen 03:03, 6 September 2015 (UTC) Dana, you're right. This was a bad mistake. I think it should be restored at his #5 state. I think there is a simple way to do this, by clicking on #5 and then "Replace the current A075783 with this version." Let me see with Joerg if this is it. Thanks. --Michel Marcus 05:12, 6 September 2015 (UTC)


I'm not sure if I've found a bug in the algorithm for the minimal surface area of a rectangular solid. Try n = 1332. s1_0 = 11, as the floor of the root. Then, we step down to s1 = 9 as its the largest divisor of 1332 under the cube root. The remaining factors' product is 148. Then s2_0 = 12 as the floor of that square root. We step down to 4 to find a divisor of 148. Thus we obtain a 9*4*37 solid, with surface area 1034. However, a solid of 6*6*37 has a surface area of 960. I believe this similar thing happens at n=68 and n=74634... potentially others. It appears that when the first divisor is too large it "conceals" a potential square or near square. I wrote more detail here: and there's a lot of math teachers discussing a problem built around this here: I wonder if I've made an error, or if the algorithm is not correct. --Scott B. Farrar 16:50, 29 September 2015 (UTC) Hi Scott, thanks for your message. Will come back you. Cheers. --Michel Marcus 17:33, 29 September 2015 (UTC)

2d-equivalent of A085690

In the discussion of A085690 you had been asking "which sequence is the 2-D equivalent" of A085690. When running the "2-d reduced program", I just get 4*n, i.e. A008586. However I'm hesitating to decorate A008586 with something like "number of circle/square intersections ..." in analogy to the 3-D case.--Hugo Pfoertner 20:28, 9 April 2016 (UTC)

A274142 etc.

I just found an optimization which allows me to generate the sequences up to the limit (1000 digits), so please do not remove the 'more' tag for the moment. I cannot submit the new files now because I already have 7 pending proposals. --Kenny Lau 13:08, 1 July 2016 (UTC)

Also, is there any way to have more than 7 pending proposals? --Kenny Lau 13:09, 1 July 2016 (UTC)

How will that limit be increased? --Kenny Lau 14:57, 1 July 2016 (UTC)


In fact, I could generate as many terms as I like (and none of the terms will exceed 1000 digits). In this case, what would the upper limit be? I mean, how many terms should I generate? --Kenny Lau 14:52, 1 July 2016 (UTC)

by mail

I notice that the "from" address is which I why I did not reply by mail. Would you receive messages if I send to ? --Kenny Lau 14:58, 1 July 2016 (UTC)

Personal tools