[seqfan] automated discovery of unknown formulae

Max Alekseyev maxale at gmail.com
Tue Oct 25 15:58:20 CEST 2011

My recent finding of a simple formula for sequence A007987 in terms of
A168597 or A002426 (while the sequence was originated from a very
complex combinatorial formula taking a whole page in Humphries 1997
paper) suggested that the following project may greatly benefit our

Run Superseeker on a dedicated server (with a local copy of the OEIS)
on every sequence Axxxxxx and report matches with a sequence Ayyyyyy

(i) the match is conservative: say, both Axxxxxx and Ayyyyyy are long
enough and T(Axxxxxx), where T is a matching transformation, matches
all known terms of Ayyyyyy.

(ii) a reference to Ayyyyyy is missing in the OEIS entry of Axxxxxx
(or more relaxed variation: Ayyyyyy is not referenced from the formula
field of the OEIS entry of Axxxxxx)

This would find a lot of previously unknown (conjectured) formulae for
sequences in terms of other sequences in the OEIS.
Of course, there would likely be many false hits but hopefully we
would be able to filter them by manual inspection.

This project would easily found a "my" formula for A007987 (and
hopefully many many others).

Is anybody interested in setting up such a project?


More information about the SeqFan mailing list