[Paper Review] On the multidimensional permanent and q-ary designs

Vladimir N. Potapov|arXiv (Cornell University)|Jan 19, 2011

graph theory and CDMA systems9 references3 citations

TL;DR

This paper introduces and analyzes two generalizations of combinatorial designs—H-designs and A-designs—within the framework of q-ary hypercubes, linking their existence and enumeration to the multidimensional permanent. It proves the existence of H(2^{t+1}, s2^t, 2^{t+1}-1, 2^{t+1}-2) designs for all s,t ≥ 1 and establishes that the number of such designs corresponds to the multidimensional permanent of an adjacency array derived from a k-partite hypergraph structure.

ABSTRACT

An $H(n,q,w,t)$ design is considered as a collection of $(n-w)$-faces of the hypercube $Q^n_q$ perfectly piercing all $(n-t)$-faces. We define an $A(n,q,w,t)$ design as a collection of $(n-t)$-faces of hypercube $Q^n_q$ perfectly cowering all $(n-w)$-faces. The numbers of H- and A-designs are expressed in terms of multidimensional permanent. We present several constructions of H- and A-design and prove the existence of $H(2^{t+1},s2^t,2^{t+1}-1,2^{t+1}-2)$ designs for every $s,t\geq 1$. Keywords: perfect matching, clique matching, permanent, MDS code, generalized Steiner system, H-design.

Motivation & Objective

• To generalize Steiner systems and t-designs into q-ary hypercube frameworks using H-designs and A-designs.
• To establish a connection between the existence and enumeration of these designs and the multidimensional permanent of adjacency arrays.
• To provide constructive methods for generating H- and A-designs, particularly for specific parameter sets.
• To investigate the relationship between these designs and known combinatorial objects such as MDS codes, perfect matchings, and clique matchings.
• To determine conditions under which partitions into H-designs or A-designs exist, especially in relation to code distance and hypergraph structure.

Proposed method

• Models H(n,q,w,t) and A(n,q,w,t) designs as subsets of codewords in Q_{q*}^n with specified weights, where * denotes missing coordinates.
• Defines H-designs as collections of (n-w)-faces that pierce each (n-t)-face exactly once, and A-designs as collections of (n-t)-faces that cover each (n-w)-face exactly once.
• Represents the incidence structure between (n-w)-faces and (n-t)-faces as a k-partite hypergraph G(n,q,w,t), with parts L = Q_q^n(w) and R = Q_q^n(t).
• Constructs an adjacency array M(G,L) or M(G,R) encoding the number of incidences between faces, and uses the k-dimensional permanent to count valid designs.
• Applies Proposition 4 and 5 to show that the number of (R,L)-perfect codes (i.e., A-designs) equals the k-dimensional permanent of M(G,L), and similarly for H-designs.
• Uses hypergraph perfect matchings and diagonal sums in the adjacency array to formalize the counting mechanism via the multidimensional permanent.

Experimental results

Research questions

• RQ1 For which parameters n, q, w, t does an H(n,q,w,t) design exist?
• RQ2 How can the number of A(n,q,w,t) designs be enumerated using combinatorial invariants like the multidimensional permanent?
• RQ3 What is the structural relationship between H-designs, A-designs, and known objects such as MDS codes and perfect matchings?
• RQ4 Under what conditions does a partition of the set of (n-w)-faces into H-designs exist?
• RQ5 Can the multidimensional permanent be used to prove the existence of H-designs for infinite families of parameters?

Key findings

• The number of A(n,q,w,t) designs is equal to the k-dimensional permanent of the adjacency array M(G(n,q,w,t), L), where k = (n-t choose n-w) * q^{w-t}.
• The number of H(n,q,w,t) designs is equal to the m-dimensional permanent of M(G(n,q,w,t), R), where m = (w choose t).
• An H(n,q,n,t) design corresponds exactly to an MDS code in Q_q^n with minimum distance d = n - t + 1.
• An A(n,q,n,n-1) design corresponds to a perfect clique matching in Q_q^n, and such designs with Hamming distance 3 exist if and only if n = 2^t.
• The existence of H(2^{t+1}, s2^t, 2^{t+1}-1, 2^{t+1}-2) designs is proven for all s,t ≥ 1.
• The construction of precise clique matchings for n = 2^{t+1}, q = 2^t is extended to a broader class of H-designs using the permanent-based counting framework.