arXiv++ Combinatorics

The Kolakoski sequence is the unique infinite sequence with values in $\{1,2\}$ and first term $1$ which equals the sequence of run-lengths of itself, we call this $K(1,2).$ We define $K(m,n)$ similarly. A well-known conjecture is that the limiting density of $K(1,2)$ is one-half. We state a natural generalization, the "generalized uniformness conjecture" (GUC). The GUC seems intractable, but we prove a partial result. The GUC implies that members of a certain family of directed graphs $G_{m,n,k}$ are all strongly connected. We prove this unconditionally. For $d>0,$ let $cf(m,n,d)$ be the density of indices $i$ such that $K(m, n)_i=K(m, n)_{i+d}.$ Essentially, $cf(m, n, d)$ is the autocorrelation function of a stationary stochastic process with random variables $\{X_t\}_{t\in\mathbb{Z}}$ whereby a sample of a finite window of this process is formed by copying as many consecutive terms of $K(m,n)$ starting from a "uniformly random index" $i\in\mathbb{Z}_+.$ Assuming the GUC, we prove that we can compute $cf(m,n,d)$ exactly for quite large $d$ by constructing a periodic sequence $S$ of period around $10^{8.5}$ such that for $d$ not too large, the correlation frequency at distance $d$ in $K(m,n)$ equals that in $S.$ We efficiently compute correlations in $S$ using polynomial multiplication via FFT. We plot our estimates $cf(m,n,d)$ for several small values of $(m,n)$ and $d\le10^5$ or $10^6$. We note many suggested patterns. For example, for the three pairs $(m,n)\in\{(1,2),(2,3),(3,4)\},$ the function $cf(m,n,d)$ behaves very differently as we restrict $d$ to the $m+n$ residue classes $\text{mod}$ $m+n.$ The plots of the three functions $cf(1,2,d),cf(2,3,d),$ and $cf(3,4,d)$ resemble waves which have common nodes. We consider this very unusual behavior for an autocorrelation function. The pairs $(m,n)\in\{(1,4),(1,6),(2,5)\}$ show wave-like patterns with much more noise.

The Kolakoski sequence is the unique infinite sequence with values in $\{1, 2\}$ and first term twems $1, 2, \ldots$ which equals the sequence of run-lengths of itself, we call this $K(1, 2).$ We define $K(m, n)$ similarly for $m+n$ odd. A well-known open problem is that its limiting density is one-half. Indeed, not much is known about the Kolakoski sequence. The focus of this paper in on conjectures related to the Kolakoski sequence which are more discrete in nature. We conjecture that a certain doubly infinite family of finite sequences $E_{1, n}\left(1^{2^j}, 1^{2j}\right)$ has odd length for all $j>0$ and even $n>0.$ We define $cf(m, n, d)$ to be the "correlation frequency" or limiting probability that terms in $K(m, n)$ which are $d$ apart are equal. We conjecture that the sign of $cf(m, n, d) - 1/2$ is periodic mod $m+n.$ We also discuss extensive empirical evidence for these conjectures.

We consider a family of integer linear programs in which the coefficients of the constraints and objective function are polynomials of an integer parameter $t.$ For $\ell$ in $\mathbb{Z}_+,$ we define $f_\ell(t)$ to be the $\ell^{\text{th}}$ largest value of the objective function with multiplicity for the integer linear program at $t.$ We prove that for all $\ell,$ $f_\ell$ is eventually quasi-polynomial; that is, there exists $d$ and polynomials $P_0, \ldots, P_{d-1}$ such that for sufficiently large $t,$ $f_\ell(t)=P_{d \pmod{t}}(t).$ Closely related to finding the $\ell^{\text{th}}$ largest value is describing the vertices of the convex hull of the feasible set. Calegari and Walker showed that if $R(t)$ is the convex hull of $\mathbf{v_1}(t), \ldots, \mathbf{v_k}(t)$ where $\mathbf{v_i}$ is a vector whose coordinates are in $\mathbb{Q}(u)$ and of size $O(u),$ then the vertices of the convex hull of the set of lattice points in $R(t)$ has eventually quasi-polynomial structure. We prove this without the $O(u)$ assumption.

The Frobenius number of relatively prime positive integers $a_1, \ldots, a_n$ is the largest integer that is not a nononegative integer combination of the $a_i.$ Given positive integers $a_1, \ldots, a_n$ with $n \ge 2,$ the set of multiples of $\gcd(a_1, \ldots, a_n)$ which have less than $m$ distinct representations as a nonnegative integer combination of the $a_i$ is bounded above, so we define $f_{m, \ell}(a_1, \ldots, a_n)$ to be the $\ell^{\text{th}}$ largest multiple of $\gcd(a_1, \ldots, a_n)$ with less than $m$ distinct representations (which generalizes the Frobenius number) and $g_m(a_1, \ldots, a_n)$ to be the number of positive multiples of $\gcd(a_1, \ldots, a_n)$ with less than $m$ distinct representations. In the parametric Frobenius problem, the arguments are polynomials. Let $P_1, \ldots, P_n$ be integer valued polynomials of one variable which are eventually positive. We prove that $f_{m, \ell}(P_1(t), \ldots, P_n(t))$ and $g_m(P_1(t), \ldots, P_n(t)),$ as functions of $t,$ are eventually quasi-polynomial. A function $h$ is eventually quasi-polynomial if there exist $d$ and polynomials $R_0, \ldots, R_{d-1}$ such that for such that for sufficiently large integers $t,$ $h(t)=R_{t \pmod{d}}(t).$ We do so by formulating a type of parametric problem that generalizes the parametric Frobenius Problem, which we call a parametric exclusion problem. We prove that the $\ell^{\text{th}}$ largest value of some polynomial objective function, with multiplicity, for a parametric exclusion problem and the size of its feasible set are eventually quasi-polynomial functions of $t.$

A $k$-majority digraph is a directed graph created by combining $k$ individual rankings on the same ground set to form a consensus where edges point in the direction indicated by a strict majority of the rankings. The $k$-majority digraph is used to model voting scenarios, where the vertices correspond to options ranked by $k$ voters. When $k$ is odd, the resulting digraph is always a tournament, called $k$-majority tournament. Let $f_k(n)$ be the minimum, over all $k$-majority tournaments with $n$ vertices, of the maximum order of an induced transitive sub-tournament. Recently, Milans, Schreiber, and West proved that $\sqrt n \le f_3(n) \le 2 \sqrt n +1 $. In this paper, we improve the upper bound of $f_3(n)$ by showing that $f_3(n) < \sqrt {2n} +\frac 12 $.