Table of Contents
Fetching ...

Relaxed vs. Full Local Decodability with Few Queries: Equivalence and Separations for Linear Codes

Elena Grigorescu, Vinayak M. Kumar, Peter Manohar, Geoffrey Mon

TL;DR

The paper investigates the boundary between relaxed locally decodable codes (RLDCs) and standard LDCs in the linear regime, proving that any linear 3-query RLDC with sufficiently good soundness is actually a 3-LDC, and formalizing a soundness threshold s(q)=2^{−⌊q/2⌋} below which a linear q-RLDC must be an LDC. It also constructs explicit small-query linear RLDCs/RLCCs (e.g., 15 and 41 queries) that are not LDCs/LCCs, establishing a separation for constant q, and extends the results to RLCCs. The core technique decomposes RLDC decoders into smooth (LDC-like) and nonsmooth parts and leverages the dual/code-structure to bound the nonsmooth component; the Goldberg transformation is refined to preserve query counts while achieving perfect completeness. The work also proves a 2-query RLDC lower bound over any finite field by showing 2-RLDCs must essentially be 2-LDCs, yielding exponential lower bounds, and provides parallel results for RLCCs. Overall, the results sharpen the landscape of local decodability vs relaxation, revealing a threshold window (between 4 and 15 for linear codes) where RLDC/RLCC equivalence to LDCs breaks down, and elucidating the role of strong soundness in this dichotomy.

Abstract

A locally decodable code (LDC) $C \colon \{0,1\}^k \to \{0,1\}^n$ is an error-correcting code that allows one to recover any bit of the original message with good probability while only reading a small number of bits from a corrupted codeword. A relaxed locally decodable code (RLDC) is a weaker notion where the decoder is additionally allowed to abort and output a special symbol $\bot$ if it detects an error. For a large constant number of queries $q$, there is a large gap between the blocklength $n$ of the best $q$-query LDC and the best $q$-query RLDC. Existing constructions of RLDCs achieve polynomial length $n = k^{1 + O(1/q)}$, while the best-known $q$-LDCs only achieve subexponential length $n = 2^{k^{o(1)}}$. On the other hand, for $q = 2$, it is known that RLDCs and LDCs are equivalent. We thus ask the question: what is the smallest $q$ such that there exists a $q$-RLDC that is not a $q$-LDC? In this work, we show that any linear $3$-query RLDC is in fact a $3$-LDC, i.e., linear RLDCs and LDCs are equivalent at $3$ queries. More generally, we show for any constant $q$, there is a soundness error threshold $s(q)$ such that any linear $q$-RLDC with soundness error below this threshold must be a $q$-LDC. This implies that linear RLDCs cannot have "strong soundness" -- a stricter condition satisfied by linear LDCs that says the soundness error is proportional to the fraction of errors in the corrupted codeword -- unless they are simply LDCs. In addition, we give simple constructions of linear $15$-query RLDCs that are not $q$-LDCs for any constant $q$, showing that for $q = 15$, linear RLDCs and LDCs are not equivalent. We also prove nearly identical results for locally correctable codes and their corresponding relaxed counterpart.

Relaxed vs. Full Local Decodability with Few Queries: Equivalence and Separations for Linear Codes

TL;DR

The paper investigates the boundary between relaxed locally decodable codes (RLDCs) and standard LDCs in the linear regime, proving that any linear 3-query RLDC with sufficiently good soundness is actually a 3-LDC, and formalizing a soundness threshold s(q)=2^{−⌊q/2⌋} below which a linear q-RLDC must be an LDC. It also constructs explicit small-query linear RLDCs/RLCCs (e.g., 15 and 41 queries) that are not LDCs/LCCs, establishing a separation for constant q, and extends the results to RLCCs. The core technique decomposes RLDC decoders into smooth (LDC-like) and nonsmooth parts and leverages the dual/code-structure to bound the nonsmooth component; the Goldberg transformation is refined to preserve query counts while achieving perfect completeness. The work also proves a 2-query RLDC lower bound over any finite field by showing 2-RLDCs must essentially be 2-LDCs, yielding exponential lower bounds, and provides parallel results for RLCCs. Overall, the results sharpen the landscape of local decodability vs relaxation, revealing a threshold window (between 4 and 15 for linear codes) where RLDC/RLCC equivalence to LDCs breaks down, and elucidating the role of strong soundness in this dichotomy.

Abstract

A locally decodable code (LDC) is an error-correcting code that allows one to recover any bit of the original message with good probability while only reading a small number of bits from a corrupted codeword. A relaxed locally decodable code (RLDC) is a weaker notion where the decoder is additionally allowed to abort and output a special symbol if it detects an error. For a large constant number of queries , there is a large gap between the blocklength of the best -query LDC and the best -query RLDC. Existing constructions of RLDCs achieve polynomial length , while the best-known -LDCs only achieve subexponential length . On the other hand, for , it is known that RLDCs and LDCs are equivalent. We thus ask the question: what is the smallest such that there exists a -RLDC that is not a -LDC? In this work, we show that any linear -query RLDC is in fact a -LDC, i.e., linear RLDCs and LDCs are equivalent at queries. More generally, we show for any constant , there is a soundness error threshold such that any linear -RLDC with soundness error below this threshold must be a -LDC. This implies that linear RLDCs cannot have "strong soundness" -- a stricter condition satisfied by linear LDCs that says the soundness error is proportional to the fraction of errors in the corrupted codeword -- unless they are simply LDCs. In addition, we give simple constructions of linear -query RLDCs that are not -LDCs for any constant , showing that for , linear RLDCs and LDCs are not equivalent. We also prove nearly identical results for locally correctable codes and their corresponding relaxed counterpart.

Paper Structure

This paper contains 30 sections, 19 theorems, 28 equations, 2 algorithms.

Key Result

Theorem 1

Let $C \colon \{0,1\}^k \to \{0,1\}^n$ be a linear $(3, \delta, 1, \frac{1}{2} - \eta)$-RLDC with a possibly adaptive decoder. Then, for any $\varepsilon > 0$, $C$ is a linear $(3, 2 \eta \delta \varepsilon/3, 1, \varepsilon)$-LDC. In particular, if $C$ is a linear $(3, \Theta(1), 1, \frac{1}{3})$-R

Theorems & Definitions (59)

  • Definition 1.4: Binary locally decodable codes; see \ref{['def:ldc']}
  • Definition 1.5: Binary relaxed locally decodable codes; see \ref{['def:rldc']}
  • Theorem 1
  • Corollary 1.6
  • Theorem 2: Soundness error threshold for $q$-RLDCs; binary case of \ref{['thm:formalrldcsoundnessthreshold']}
  • Definition 1.7: Strong soundness for LDCs and RLDCs
  • Corollary 1.8: RLDCs with Strong Soundness are LDCs
  • Theorem 3: Binary case of \ref{['thm:stronger-goldberg']}
  • Corollary 1.9
  • Theorem 4: Constructions of constant query RLDCs/RLCCs that are not LDCs
  • ...and 49 more