Stable complete coordinates for multisets of points via basic $r$-symmetric tropical polynomials
2026-06-29 • Discrete Mathematics
Discrete Mathematics
AI summaryⓘ
The authors study ways to uniquely identify unordered collections of points in multi-dimensional space that ignore the order of points. They provide a clear and efficient set of mathematical tools (called tropical polynomials) that can distinguish these collections for any number of points and dimensions, improving on earlier less efficient methods. Their approach uses ideas from a transportation problem and includes an explicit method to recover the original point sets. They also analyze the stability of their method and pinpoint exactly when simpler tools are enough and when more complex ones are necessary.
point cloudsymmetric grouporbit spacetropical polynomialsmax-plus algebrapermutation invariantsbi-Lipschitz embeddingtransportation problemdiscrete tomographypersistence barcode
Authors
Susumu Kubo
Abstract
A multiset of $n$ unordered points in $\mathbb{R}^r$ -- a point cloud, or, for $r=2$, a persistence barcode of birth-death pairs -- is a point of the orbit space $\mathbb{R}^{nr}/S_n$ for the symmetric group $S_n$ permuting the rows of an $n \times r$ matrix; a separating family of invariants on this space is exactly a complete set of permutation-independent coordinates. We provide one that is explicit, small, and stable, in the max-plus (tropical) setting: for all $n \geq 1$ and $r \geq 1$, the $\binom{n+r}{r}$ basic $r$-symmetric tropical polynomials, of degree at most $n$, separate the orbits of $S_n$ on $\mathbb{R}^{nr}$. This settles in full a problem left open in [Kubo, J. Pure Appl. Algebra 223 (2019) 72-85], where separation was known only for $r=2$ and special cases of $r \geq 3$, and yields a family far smaller and of lower degree than the general separating sets from Derksen's recent theory of tropical invariants for permutation actions ($nr + (nr)!/n!$ invariants of degree $O(n^2 r^2)$). The proof is elementary and constructive: the basic values are identified with a transportation problem, and the multiset is recovered from the dual by an explicit algorithm. We further show the coordinate map is a bi-Lipschitz embedding for all $n$ and $r$, being an injective max filter bank (via the bi-Lipschitz theory of max filtering), with an explicit Lipschitz constant for the forward bound and a fully explicit, dimension-free distortion when $r=1$. Finally we determine when the pairwise values suffice (exactly $n \leq 3$) and show that invariants on at least three columns and of degree less than $n$ are necessary in general, the obstruction being a standard non-uniqueness configuration from discrete tomography.