32
$\begingroup$

The Hofstadter–Conway \$10000 sequence is defined by the nested recurrence relation $$c(n) = c(c(n-1)) + c(n-c(n-1))$$ with $c(1) = c(2) = 1$. This sequence is A004001 and it is well-known that this sequence has many amazing properties which are investigated in a very interesting paper Kubo and Vakil - On Conway's recursive sequence.

A related curious sequence can be defined by similar recurrence relation as below $$c^*(n) = n - c^*(c^*(n-1)) - c^*(n-c^*(n-1))$$ with $c^*(1) = c^*(2) = 1$. I introduced this sequence in A287422 and recurrence is also investigated in terms of sensitivity of initial conditions selections Alkan and Aybar - On Families of Solutions for Meta-Fibonacci Recursions Related to Hofstadter–Conway $10000 Sequence.

Conjecture. $c(n) - \frac{n}2$ $\ge$ $\left\lvert c^*(n) - \frac{n}2 \right\rvert$ for all $n \ge 1$. (It is checked up to $2^{32}$.)

Question. Can someone show that $c(n) - \frac{n}2$ $\ge$ $\left\lvert c^*(n) - \frac{n}2 \right\rvert$ for all $n \ge 1$ ?

Comments that are related to combinatorial characterization of $c^*(n)$ are also very welcome.

(I the share below graph in order to display behaviours of both sequences for $n \le 2^{10}$. Red: $c(n) - \frac{n}2$, Black: $c^*(n) - \frac{n}2$.)

Graphs of c(n) - n/2 and c*(n) - n/2

Note. I believe that it is also possible to see a way to connection of $c(n)$ and $c^*(n)$ thanks to certain auxiliary variants defined by A317754 and A317854, see the below graph (red: transformation of $c(n)$, black: transformation of $c^*(n)$). (These variants also make probably interesting sounds if one can think that recurrences that produce sequences with curious sounds are general of interest.) Graphs of transformations of c(n) and c*(n)

$\endgroup$
2
  • 3
    $\begingroup$ Has anybody tried to plot the red arcs against the properly scaled Takagi function (better still, to plot the difference between the two)? $\endgroup$
    – Seva
    Commented Dec 5, 2021 at 13:28
  • 1
    $\begingroup$ @Seva Sure enough, that readily comes to mind. On p.20-21 of the first quoted paper (excellent indeed!), a parametrization of the limit curve is developed, in terms of the erf function...! No reference to the Takagi/Blancmange curve though. $\endgroup$
    – Wolfgang
    Commented Dec 5, 2021 at 20:40

0

You must log in to answer this question.