[seqfan] Re: Breaking news on partition numbers.
charles.greathouse at case.edu
Tue Jan 25 22:28:51 CET 2011
There are two issues here: bad sequences that make it hard to search
for good sequences, and sequences that aren't fleshed out enough to be
The first problem varies (almost) proportionally with the number of
such sequences. If I had to search through ten times as many
sequences as I should have to to find the one I'm looking for, that's
five times as bad as if I had to search through twice as many. This
problem, in my view, is not a serious one: most OEIS sequences are
valuable, so the extra cost is fractional (10% more time, not 10x the
time). For example, counting the sequences Joerg references below I
find less than 600; out of nearly 200,000 this is a tiny percentage
even if it only represents a tenth of the sequences. (Of course there
are many good sequences mixed in there; I'm not going to go through
them all right now!)
The second seems more pernicious. If a person searches for a sequence
and finds it, but there is no real information present (perhaps not
even enough to determine if the sequence is in fact a match) this is a
loss. I had such an experience a time ago; that sequence has since
been fixed with the addition of formulas, comments, and further terms.
In large part I think the problems are linked. If a person submits
many sequences a day and consequently spends little time on each, then
some sequences will be uninteresting or even "noise", while the truly
good ones won't have enough information to be useful. So I often find
myself asking, "why is this interesting?". If a sequence really is
interesting, _please_ explain why so that the next person who stumbles
upon it is informed. If you find yourself writing a sequence in
significantly less time than an hour (say, just 20 minutes), stop and
ask yourself if it is worthy of inclusion.
To go back to one of Joerg's examples, the Lyman sequences: When I
look at these I can't tell what value they may have. I strongly
suspect that there are a number of sequences and comments of real
value, but since there are so many for which the value is unclear,
it's hard to tell.
Case Western Reserve University
On Tue, Jan 25, 2011 at 3:49 PM, Joerg Arndt <arndt at jjj.de> wrote:
> * Charles Greathouse <charles.greathouse at case.edu> [Jan 25. 2011 20:04]:
> <words I mark for future citation>
>> I believe in Neil's old rule of thumb here: a sequence should take at
>> least an hour to create, between research, referencing, calculating,
>> and writing.* If you can't be bothered to spend that much time, the
>> sequence might not be interesting enough for inclusion (but rather as
>> a comment to an existing sequence, perhaps).
> </words I mark for future citation>
> One thing I'd like to add:
> ... drum-roll ...
> *** tah-dah! ***
> <more words to mark, oh, what a day>
> Anti-information is bad. It reduces the usefulness of the OEIS.
> </more words to mark, oh, what a day>
> Anti-information techniques that make my blood boil, just from the top
> of my head (no order, some redundant redundancy):
> - random mix of number theoretic functions
> - unranking, mixed unranking of _different_ things
> - unranking and plain mixed (argh!)
> - unranking mixed with ranking, same with different things for extra shock value
> - seq of indices "where seq. S has values <random function>"
> - sum of a (linear) sequence with a triangle or rectangle, hey, all three types at once
> - linear transform with a triangle/rectangle
> - fractions of utterly disconnected seqs
> - single integer, followed by sequence S, no info but the bare terms in it, few of them
> - twice sequence S, thrice S, etc., only few terms for neat uselessness
> - first differences of the most contrived ever seqs
> - second diffs as well, ...we are in the mood
> - playing with primes, semi-primes, composites
> - mixing stuff of the last item
> - ... adding unranking for complete brain explosion's sake
> - ad hoc terminology (extra points for a few syntax failures)
> - last item, _nowhere_ bloody explained
> - "you must read my mind" approach: "like", "similar", "almost", "related to"
> - giving <=10 terms of a seq where 1000 terms can be computed in < 1 sec
> - garbage seq with link to super nuclear garbage web document, MS Word for extra credit
> - Calling it "Generalized X" for X in Fibonacci/Catalan/Pascal/you-name-it
> - crossreffing A000079 because one longish formula involves ... powers of 2
> - crossreffing a constant sequence for a constant in some formula
> - long bleeding monologues you know this is fun to read and then I found 5 dollars (see A010716).
> - last item, and saying "quantum", "fractal" and the like but not defining your sequence.
> - calling the not-so-atrociously-restricted version of <your-ad-hoc-term>
> the "generalized <your-ad-hoc-term>" (I. Must. Kill. Arrrrr...)
> The following searches will give you some illustrative examples,
> obviously together with plenty of fine seqs (the vast majority,
> just to be clear). One line is one search term:
> "matrix Markov"
> "generalized Fibonacci"
> keyword:easy keyword:more
> cheers, jj
> Is there a chainsaw over internet protocol?
> Seqfan Mailing list - http://list.seqfan.eu/
More information about the SeqFan