Hello SeqFan,
For
A393385 - OEIS "
number of minimal edge covers in the n-necklace graph", there is a question about what to do with a(0).
For n >= 1, the counts begin:
7, 69, 571, 4729, 39227
A transfer-matrix/trace derivation leads to the g.f.:
(3 - 14*x - 10*x^2)/(1 - 7*x - 10*x^2 - 6*x^3)
If you try and interpret n = 0 combinatorially as the "empty necklace" (the empty graph), there is exactly 1 minimal edge cover (the empty set) that corresponds to the normalised g.f.:
(1 + 10*x^2 + 12*x^3)/(1 - 7*x - 10*x^2 - 6*x^3)
which differs from the trace g.f. by a constant (normalised = trace - 2), and therefore leaves all coefficients for n >= 1 unchanged.
Question: For cyclically-constructed graph families that are only canonically defined for n >= 1, is it better to (a) keep the recurrence/trace continuation a(0) = 3, (b) normalised so that a(0) = 1 matches the "empty object" convection, or (c) set OFFSET 1 and avoid n = 0 entirely?