[Paper Review] Subspace Polynomials and Cyclic Subspace Codes

Eli Ben‐Sasson, Tuvi Etzion+2 more|arXiv (Cornell University)|Apr 30, 2014

Coding theory and cryptography18 references4 citations

TL;DR

This paper introduces subspace polynomials as a novel algebraic tool to construct cyclic subspace codes with optimal parameters. It proves the existence of cyclic codes in $\mathcal{G}_q(n,k)$ with minimum distance $2k-2$ for infinitely many $n$ when $k$ is fixed, using field extension embeddings and inclusion-exclusion on orbit structures, and provides explicit constructions for codes with full or degenerate orbits.

ABSTRACT

Subspace codes have received an increasing interest recently due to their application in error-correction for random network coding. In particular, cyclic subspace codes are possible candidates for large codes with efficient encoding and decoding algorithms. In this paper we consider such cyclic codes and provide constructions of optimal codes for which their codewords do not have full orbits. We further introduce a new way to represent subspace codes by a class of polynomials called subspace polynomials. We present some constructions of such codes which are cyclic and analyze their parameters.

Motivation & Objective

• To address the lack of general constructions for cyclic subspace codes with full or degenerate orbits, particularly those with optimal minimum distance.
• To develop a new representation of subspaces using subspace polynomials that enables efficient analysis of subspace intersections and orbit structures.
• To prove the existence of cyclic codes with minimum distance $2k-2$ in $\mathcal{G}_q(n,k)$ for infinitely many $n$ when $k$ is fixed.
• To explore the embedding of cyclic codes over smaller fields into larger Grassmannians while preserving cyclicity and scaling minimum distance.
• To provide explicit constructions of cyclic codes with full-length orbits and analyze their automorphism groups, particularly when $n$ is prime.

Proposed method

• Represent subspaces via subspace polynomials—specialized linearized polynomials whose roots span the subspace—enabling algebraic manipulation of subspace properties.
• Use the concept of gap between two polynomials to bound the dimension of intersection of their corresponding subspaces, linking polynomial structure to subspace geometry.
• Apply the inclusion-exclusion principle to compute the size of a cyclic code formed by combining subcodes associated with divisors of $n$, based on orbit decomposition.
• Embed cyclic codes from $\mathcal{G}_{q^d}(n/d, k/d)$ into $\mathcal{G}_q(n,k)$ via field norm maps, preserving cyclicity and scaling minimum distance by a factor of $d$.
• Leverage the structure of Singer subgroups and Frobenius automorphisms to show that the automorphism group of the constructed codes contains the normalizer of a Singer subgroup.
• Use the fact that $\text{gap}(P_1, P_2) \geq d$ implies $\dim(V_1 \cap V_2) \leq \dim(V_1) - d/2$, which helps bound the minimum distance of the code.

Experimental results

Research questions

• RQ1 Can cyclic subspace codes with minimum distance $2k-2$ be constructed explicitly for infinitely many $n$ when $k$ is fixed?
• RQ2 How can subspace polynomials be used to represent and analyze the structure of cyclic subspace codes, particularly their orbits and intersections?
• RQ3 What is the relationship between the orbit structure of a subspace and the algebraic properties of its associated subspace polynomial?
• RQ4 Can cyclic codes over smaller fields be embedded into larger Grassmannians while preserving cyclicity and controlling minimum distance?
• RQ5 Is it possible to construct cyclic codes with multiple orbits, especially when the orbits are degenerate (i.e., smaller than $\frac{q^n-1}{q-1}$)?

Key findings

• The paper proves that for any fixed $k$ and infinitely many $n$, there exist cyclic subspace codes in $\mathcal{G}_q(n,k)$ with minimum distance $2k-2$, resolving a long-standing conjecture in special cases.
• Explicit constructions of such codes are provided using subspace polynomials and orbit decomposition, with the size of the code given by an inclusion-exclusion formula over divisors of $n$.
• When $n$ is prime, the constructed codes have automorphism group containing the normalizer of a Singer subgroup, which is a strong structural property.
• The size of a cyclic code formed by subcodes corresponding to divisors $d_i$ of $n$ is given by $|\mathbb{C}| = \sum_{i} {n/d_i \brack k/d_i}_{q^{d_i}} - \sum_{i<j} {n/\text{lcm}(d_i,d_j) \brack k/\text{lcm}(d_i,d_j)}_{q^{\text{lcm}(d_i,d_j)}} + \cdots$, using inclusion-exclusion.
• A cyclic code $\mathbb{C} \subseteq \mathcal{G}_{q^d}(n/d, k/d)$ with minimum distance $2(k/d) - 2\delta$ can be embedded into $\mathcal{G}_q(n,k)$ as $\mathbb{C}'$, preserving size and yielding minimum distance $2k - 2d\delta$.
• The paper shows that the gap between two subspace polynomials provides a lower bound on the minimum distance of the corresponding code, though it may not be the most efficient measure for intersection size.