[Paper Review] A New Construction for Constant Weight Codes

Tuvi Etzion, Alexander Vardy|arXiv (Cornell University)|Apr 9, 2010

graph theory and CDMA systems39 references16 citations

TL;DR

This paper presents a novel construction—FDTW—for constant weight codes by mapping $k$-dimensional subspaces from the Grassmannian space $\mathcal{G}_q(n,k)$ into binary constant weight codewords of length $q^n$ and weight $q^k$. The method leverages finite field representations and coset structures, yielding new optimal or largest-known constant weight codes with efficient encoding, decoding, and error-correction algorithms, particularly effective under bounded error rates.

ABSTRACT

A new construction for constant weight codes is presented. The codes are constructed from $k$-dimensional subspaces of the vector space $\F_q^n$. These subspaces form a constant dimension code in the Grassmannian space $\cG_q(n,k)$. Some of the constructed codes are optimal constant weight codes with parameters not known before. An efficient algorithm for error-correction is given for the constructed codes. If the constant dimension code has an efficient encoding and decoding algorithms then also the constructed constant weight code has an efficient encoding and decoding algorithms.

Motivation & Objective

• To develop a new method for constructing constant weight codes with improved or previously unknown parameters.
• To establish a systematic link between constant dimension codes in the Grassmannian and constant weight codes in binary space.
• To design efficient encoding, decoding, and error-correction algorithms for the resulting constant weight codes.
• To demonstrate that optimal or best-known constant weight codes can be derived from known constant dimension codes using the proposed construction.

Proposed method

• The construction maps each $k$-dimensional subspace $X \subseteq \mathbb{F}_q^n$ to a binary vector of length $q^n$ and weight $q^k$, representing the characteristic vector of the subspace's elements in $\mathbb{F}_{q^n}$.
• Cosets of the subspace $X$ are similarly mapped, preserving the weight and enabling the generation of multiple codewords from a single subspace.
• The construction uses the isomorphism between $\mathbb{F}_q^n$ and $\mathbb{F}_{q^n}$, mapping field elements to vectors via their $q$-ary representations.
• Error-correction is performed by analyzing the multiset of pairwise differences $\mathcal{T}(Y)$ from the received word $Y$, identifying the most frequent differences to reconstruct the original subspace.
• For $q$ even, each difference in the subspace appears $q^k/2$ times; for $q$ odd, it appears $q^k$ times, enabling robust error detection based on frequency thresholds.
• The algorithm corrects up to $\frac{q^k}{2} - \tau$ errors by identifying the most frequently used elements in $\mathcal{T}(Y)$ and recovering the original coset via the most frequent shift $\beta$.

Experimental results

Research questions

• RQ1 Can constant dimension codes in the Grassmannian $\mathcal{G}_q(n,k)$ be systematically transformed into constant weight codes with improved or new parameters?
• RQ2 What is the relationship between the minimum distance of the original constant dimension code and the resulting constant weight code?
• RQ3 Can efficient encoding and decoding algorithms be inherited from the constant dimension code to the constant weight code?
• RQ4 What error-correction capability does the constructed constant weight code achieve, and how does it depend on the field size $q$ and dimension $k$?
• RQ5 Are there new optimal constant weight codes with parameters not previously known, derivable via this construction?

Key findings

• The construction produces constant weight codes with parameters $ (q^n, 2q^k - 2q^{k-t}, q^k) $, where $ t $ is related to the minimum distance of the original constant dimension code.
• Several of the constructed codes are optimal or the largest known for their parameters, including new optimal codes not previously documented.
• The error-correction algorithm can correct any number of errors less than $ \frac{q^k}{2} $, based on frequency analysis of pairwise differences in the received word.
• The method ensures that if the original constant dimension code has efficient encoding/decoding, so does the resulting constant weight code.
• For $q$ even, each difference in a codeword appears $ \frac{q^k}{2} $ times in $ \mathcal{T}(X) $; for $q$ odd, it appears $ q^k $ times, enabling reliable reconstruction under bounded errors.
• The construction generalizes prior constructions and includes them as special cases, demonstrating broader applicability and improved code size.