[seqfan] Re: LRS formulas for n>=...?

[rhhardin at att.net]

Of course that automatic detection is what I would be doing, if I included
the recurrence, because I detect it automatically as well.  (The actual
calculation for my series is essentially brute force enumeration.)

But surely unverified truth turns up all over; somebody enumerates out
a series, and somebody else comes up with the same series on a different
problem, and so adds a comment rather than submit his series, even if he
has fewer terms; or he may add terms as well as the comment.

Now there are two problems with the same series, and the series is extended
by either problem, without anybody checking it applies for the other.

I think you have to just judge the odds: how many terms with how many
digits, and what are the chances this is accidental.

If your judgment is good, it works for both problems.

The real difficulty is can you automatically judge the chances.
--
rhhardin at mindspring.com
rhhardin at att.net (either)
  
-------------- Original message ----------------------
From: Richard Mathar <mathar at strw.leidenuniv.nl>
> As the database is supposed to contain the verified truth, it's a scaring
> option to automate the detection. One could in some way collect
> all the apparent recurrences by running some sort of superseeker across the
> entire database and put them in some index, but that would remain labeled
> by "use with caution".