[Paper Review] Constructions of Snake-in-the-Box Codes under $\ell_{\infty}$-metric for Rank Modulation

Xiang Wang, Fang‐Wei Fu|arXiv (Cornell University)|Jan 21, 2016

Coding theory and cryptography13 references3 citations

TL;DR

This paper presents two novel constructions of $ε_{\infty}$-snakes—longest possible Gray codes under the $ε_{\infty}$-metric in the rank modulation scheme—for flash memory error detection. The first method uses cyclic and complete rank modulation Gray codes (RMGCs), while the second leverages $Κ$-snakes to achieve longer codes. The key contribution is a significant improvement in code length over prior work, with the second construction yielding up to 342 codewords for $n=7$, surpassing previous bounds by 126.

ABSTRACT

In the rank modulation scheme, Gray codes are very useful in the realization of flash memories. For a Gray code in this scheme, two adjacent codewords are obtained by using one "push-to-the-top" operation. Moreover, snake-in-the-box codes under the $\ell_{\infty}$-metric are Gray codes, which can be capable of detecting one $\ell_{\infty}$-error. In this paper, we give two constructions of $\ell_{\infty}$-snakes. On the one hand, inspired by Yehezkeally and Schwartz's construction, we present a new construction of the $\ell_{\infty}$-snake. The length of this $\ell_{\infty}$-snake is longer than the length of the $\ell_{\infty}$-snake constructed by Yehezkeally and Schwartz. On the other hand, we also give another construction of $\ell_{\infty}$-snakes by using $\mathcal{K}$-snakes and obtain the longer $\ell_{\infty}$-snakes than the previously known ones.

Motivation & Objective

• To develop longer $ε_{\infty}$-snake codes under the rank modulation scheme for improved error detection in flash memory.
• To overcome the limitations of existing constructions, which produce shorter codes than the theoretical upper bound.
• To introduce two new methods—using cyclic and complete RMGCs and $Κ$-snakes—that generate longer $ε_{\infty}$-snakes than prior constructions.
• To establish a framework for constructing $ε_{\infty}$-snakes with guaranteed minimum distance 2, enabling single-error detection.
• To improve upon Yehezkeally and Schwartz’s construction by achieving larger code lengths for $n \geq 6$ and specific $n=4k\pm1$.

Proposed method

• The first construction uses cyclic and complete rank modulation Gray codes (RMGCs) to generate $ε_{\infty}$-snakes, leveraging transition sequences that ensure adjacent codewords differ by a single 'push-to-the-top' operation.
• The second construction employs $Κ$-snakes—specifically, a $k$-snake structure in alternating groups $A_n$—to build longer $ε_{\infty}$-snakes by concatenating transition sequences.
• The method applies a recursive structure: codewords are generated by applying a sequence of 'push-to-the-top' operations, ensuring each step produces a new permutation in $S_n$.
• The length of the resulting $ε_{\infty}$-snake is derived from the product of factorial terms and combinatorial counts: $M_{n,1} = \lceil n/2 \rceil! (\lfloor n/2 \rfloor + (\lfloor n/2 \rfloor)! )$ for $n \geq 6$, and $M_{n,2}$ for $n=4k\pm1$.
• For $n=7$, the second construction yields a cyclic $(7,342,\ell_{\infty})$-snake, confirmed via explicit construction in Figure 5 using $\hat{\mathcal{T}}_{\mathcal{K},5}$ and $\mathcal{T}_3$.
• Theoretical bounds are derived using group-theoretic properties of $S_n$ and $A_n$, ensuring that adjacent codewords differ by at most one transposition under the $ε_{\infty}$-metric.

Experimental results

Research questions

• RQ1 Can we construct $ε_{\infty}$-snakes in $S_n$ that are longer than the construction by Yehezkeally and Schwartz?
• RQ2 What structural properties of RMGCs and $Κ$-snakes enable the construction of longer $ε_{\infty}$-snakes under the $ε_{\infty}$-metric?
• RQ3 Is there a systematic method to generate $ε_{\infty}$-snakes with length exceeding $\lceil n/2 \rceil! (\lfloor n/2 \rfloor + (\lfloor n/2 \rfloor - 1)! )$ for $n \geq 6$?
• RQ4 For which values of $n$ does the $Κ$-snake-based construction outperform the RMGC-based construction in terms of code length?
• RQ5 Can cyclic and complete RMGCs be used to generate complete, non-cyclic $ε_{\infty}$-snakes with maximal length for $n \geq 6$?

Key findings

• The first construction produces an $(n, M_{n,1}, \ell_{\infty})$-snake of length $M_{n,1} = \lceil n/2 \rceil! (\lfloor n/2 \rfloor + (\lfloor n/2 \rfloor)! )$ for all $n \geq 6$, which is strictly longer than Yehezkeally and Schwartz’s construction.
• For $n=7$, the first construction yields $M_{7,1} = 216$ codewords, exceeding the prior bound of $M_{7,0} = 120$.
• The second construction using $\mathcal{K}$-snakes achieves $M_{n,2} = \left(\frac{(2k+1)!}{2} - 2k + 1\right) \cdot (2k+1)!$ for $n=4k+1$, and a similar expression for $n=4k-1$, with $k \geq 2$, yielding longer codes than both prior constructions.
• For $n=7$ ($k=2$), the $\mathcal{K}$-snake method produces a cyclic $(7,342,\ell_{\infty})$-snake, which is the longest known at this time.
• The $\mathcal{K}$-snake construction surpasses both the RMGC-based and Yehezkeally-Schwartz constructions for $n=4k+1$ or $n=4k-1$ with $k \geq 2$, as shown by $M_{n,2} > M_{n,1} > M_{n,0}$.
• The paper confirms the existence of a cyclic and complete $3$-RMGC with transition sequence $\mathcal{T}_3 = (t_3,t_3,t_2,t_3,t_3,t_2)$, which is used as a building block in the $\mathcal{K}$-snake construction for $n=7$.