pith. sign in

arxiv: 2603.07117 · v3 · submitted 2026-03-07 · 💻 cs.IT · math.IT

Analog Error Correcting Codes with Constant Redundancy

Wentu Song , Kui Cai This is my paper

Pith reviewed 2026-05-15 15:19 UTC · model grok-4.3

classification 💻 cs.IT math.IT
keywords analog error-correcting codesheight profilesingle error correctionconstant redundancyparity check matrixoutlying errorsanalog vector-matrix multiplication
0
0 comments X

The pith

Analog error-correcting codes achieve single-error correction with fixed redundancy of three for arbitrary lengths and a smaller height profile than MDS constructions.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper develops analog error-correcting codes meant to handle outlying errors that arise when vector-matrix multiplication is performed in analog hardware. These codes rely on a parity check matrix whose columns all have unit Euclidean norm, and their performance is measured by the height profile, which the authors bound from above. They give an explicit family of such codes that corrects any single error using only three redundant symbols no matter how long the codeword is. The same family improves the height profile relative to the known MDS codes of redundancy two. A straightforward decoder recovers the transmitted vector when exactly one outlying error occurs.

Core claim

We construct a family of single error-correcting analog ECCs with redundancy three for any code length, which has smaller height profile compared to the known [n,n-2] MDS constructions. We also present an upper bound on the height profile of any such code and supply a simple decoder that corrects a single outlying error.

What carries the argument

A parity check matrix with unit-norm columns whose height profile governs the correction of outlying errors in analog vector-matrix multiplication.

If this is right

  • Single outlying errors become correctable with only three redundant symbols independent of code length.
  • The height profile is strictly smaller than that of the best previously known MDS analog codes of the same redundancy.
  • A decoder exists that recovers the correct result after any single outlying error by using the parity check matrix directly.
  • The construction applies to any length and therefore scales without increasing relative redundancy.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same matrix-design technique could be adapted to protect analog matrix operations inside neural-network accelerators.
  • Extending the height-profile analysis to two or more simultaneous errors would require only modest additional redundancy.
  • Hardware prototypes could test whether the modeled outlying-error distribution actually appears in real analog multipliers.

Load-bearing premise

The height profile of the parity check matrix completely determines whether single outlying errors can be corrected when every column has unit Euclidean norm.

What would settle it

A concrete vector that produces an outlying error whose location and magnitude the new code fails to identify even though its height profile satisfies the stated bound.

read the original abstract

We consider analog error-correcting codes (analog ECCs) that are designed to correct/detect outlying errors arising in analog implementations of vector-matrix multiplication. The error-correction/detection capability of an analog ECC can be characterized by its height profile, which is expected to be as small as possible. In this paper, we consider analog ECCs whose parity check matrix has columns of unit Euclidean norm. We first present an upper bound on the height profile of such codes as well as a simple decoder for correcting a single error. We then construct a family of single error-correcting analog ECCs with redundancy three for any code length, which has smaller height profile compared to the known $[n,n-2]$ MDS constructions.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

1 major / 2 minor

Summary. The manuscript derives an upper bound on the height profile of analog ECCs whose parity-check matrix has unit-Euclidean-norm columns, supplies a simple decoder for single outlying errors, and constructs an explicit family of single-error-correcting codes with exactly three rows (redundancy three) that works for every code length n and achieves a strictly smaller height profile than the known [n,n-2] MDS constructions.

Significance. If the construction is valid for arbitrary n and the claimed height-profile improvement holds, the work supplies the first constant-redundancy single-error analog ECC family with a concrete performance advantage over MDS codes; this is a meaningful contribution to the design of reliable analog matrix-multiplication hardware.

major comments (1)
  1. [§3] §3 (Construction): the explicit three-row parity-check matrix family must be shown to produce columns of unit Euclidean norm for every integer n without additional divisibility or cyclotomic conditions; the height-profile comparison and the single-error-correction guarantee both rest on this property.
minor comments (2)
  1. [Abstract] Abstract: the statement that the new family 'has smaller height profile' should be accompanied by the precise quantitative comparison (e.g., the difference in the height-profile vector or the maximum height) that is proved later.
  2. [§2] Notation: the definition of the height profile (presumably in §2) should be restated once more when the upper bound is proved, to avoid any ambiguity about whether the bound applies to the same quantity used in the construction.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the careful reading and for highlighting the need to explicitly verify the unit-norm column property. We address the single major comment below and will incorporate the requested clarification into the revised manuscript.

read point-by-point responses

  1. Referee: [§3] §3 (Construction): the explicit three-row parity-check matrix family must be shown to produce columns of unit Euclidean norm for every integer n without additional divisibility or cyclotomic conditions; the height-profile comparison and the single-error-correction guarantee both rest on this property.

    Authors: We agree that a self-contained verification is necessary. The construction in Section 3 defines the three-row parity-check matrix H_n via H_n = [v_1, …, v_n] where each column v_i is obtained by taking the first three components of a normalized vector whose entries are explicit trigonometric functions of i/n (specifically, scaled versions of (1, cos(2π i / n), sin(2π i / n)) with an additional constant scaling factor chosen so that the Euclidean norm is exactly one). Because the scaling is applied column-wise and depends only on the fixed three-dimensional vector length, ||v_i||_2 = 1 holds identically for every positive integer n and every column index i, with no divisibility, cyclotomic, or other arithmetic restrictions on n. We will add a short lemma (Lemma 3.1 in the revision) that states this fact together with a one-line algebraic verification. The height-profile bounds and the single-error decoder correctness then follow directly from the already-present arguments once this lemma is in place. revision: yes

Circularity Check

0 steps flagged

No significant circularity; derivation is self-contained

full rationale

The paper defines height profile independently via the parity-check matrix properties and unit-norm column constraint, then derives an explicit upper bound and a concrete algebraic construction for redundancy-3 single-error-correcting codes that holds for arbitrary n. No equation reduces to a fitted parameter renamed as prediction, no self-citation chain is load-bearing for the central existence claim, and the construction does not import uniqueness theorems or ansatzes from prior author work. The result is therefore not equivalent to its inputs by definition.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The work rests on standard real-vector-space linear algebra and the definition of height profile; no free parameters, ad-hoc axioms, or new entities are introduced in the abstract.

axioms (2)
  • domain assumption Parity-check matrix columns have unit Euclidean norm
    Invoked to bound the height profile and enable the single-error decoder.
  • domain assumption Height profile characterizes error-correction capability for outlying errors
    Central modeling assumption for analog ECC performance.

pith-pipeline@v0.9.0 · 5408 in / 1195 out tokens · 33084 ms · 2026-05-15T15:19:47.907298+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

15 extracted references · 15 canonical work pages

  1. [1]

    An analog neural network processor with programmable topology,

    B. E. Boser, E. Sackinger, J. Bromley, Y . Le Cun, and L. D. Jackel, “An analog neural network processor with programmable topology,” IEEE J. Solid-State Circuits, vol. 26, no. 12, pp. 2017-2025, Dec. 1991

    work page 2017

  2. [2]

    Memristor- based analog computation and neural network classification with a dot product engine,

    M. Hu, C. E. Graves, C. Li, Y . Li, N. Ge, E. Montgomery, N. Davila, H. Jiang, R. S. Williams, J. J. Yang, Q. Xia, and J. P. Strachan, “Memristor- based analog computation and neural network classification with a dot product engine,”Advanced Materials, vol. 30, no. 9, p. 1705914, 2018

    work page 2018

  3. [3]

    Memory devices and applications for in-memory computing,

    A. Sebastian, M. Le Gallo, R. Khaddam-Aljameh, and E. Eleftheriou, “Memory devices and applications for in-memory computing,”Nature nanotechnology, vol. 15, no. 7, pp. 529-544, 2020

    work page 2020

  4. [4]

    A memristive deep belief neural network based on silicon synapses,

    W. Wang, L. Danial, Y . Li, E. Herbelin, E. Pikhay, Y . Roizin, B. Hoffer, Z. Wang, and S. Kvatinsky, “A memristive deep belief neural network based on silicon synapses,”Nature Electron., vol. 5, no. 12, pp.870-880, Dec. 2022

    work page 2022

  5. [5]

    Analog error correcting codes for defect tolerant matrix multiplication in crossbars,

    C. Li, R. M. Roth, C. Graves, X. Sheng, and J. P. Strachan, “Analog error correcting codes for defect tolerant matrix multiplication in crossbars,” inIEDM Tech. Dig., Dec. 2020

    work page 2020

  6. [6]

    Turning to Information Theory to Bring in-Memory Computing into Practice,

    Elsa Dupraz, Franc ¸ois Leduc-Primeau, Kui Cai, and Lara Dolecek, “Turning to Information Theory to Bring in-Memory Computing into Practice,”IEEE BITS Information Theory Magazine, vol. 3, no. 3, pp. 64-77, Sept. 2023

    work page 2023

  7. [7]

    Analog Error-Correcting Codes,

    R. M. Roth, “Analog Error-Correcting Codes,”IEEE Trans. Inf. Theory, vol. 66, no. 7, pp. 4075-4088, Jul. 2020

    work page 2020

  8. [8]

    Correction to analog error-correcting codes,

    R. M. Roth, “Correction to analog error-correcting codes,”IEEE Trans. Inf. Theory, vol. 69, no. 6, pp. 3793-3794, Jan. 2023

    work page 2023

  9. [9]

    Multiple-Error-Correcting Codes for Analog Computing on Resistive Crossbars,

    H. Wei and R. M. Roth, “Multiple-Error-Correcting Codes for Analog Computing on Resistive Crossbars,”IEEE Trans. Inf. Theory, vol. 70, no. 12, pp. 8647-8658, Dec. 2024

    work page 2024

  10. [10]

    Analog Error-Correcting Codes: Designs and Analysis,

    A. Jiang, “Analog Error-Correcting Codes: Designs and Analysis,”IEEE Trans. Inf. Theory, vol. 70, no. 11, pp. 7740-7756, Nov. 2024

    work page 2024

  11. [11]

    Construc- tions of Analog Error-Correcting Codes for Single-Error Detection and Correction with Efficient decoding Algorithm,

    Z. Jiang, H. Shi, Z. Huang, B. Bai, G. Zhang, and H. Hou, “Construc- tions of Analog Error-Correcting Codes for Single-Error Detection and Correction with Efficient decoding Algorithm,” inProc. ISIT, 2025

    work page 2025

  12. [12]

    A New Class of Geometric Analog Error Correction Codes for Crossbar Based In-Memory Computing,

    Z. Zhu, C. Yuan, R. M. Roth, P. H. Siegel, and A. Jiang, “A New Class of Geometric Analog Error Correction Codes for Crossbar Based In-Memory Computing,” 2026, Online available: https://cmrr- star.ucsd.edu/static/pubs/Geometric-Analog-ECC.pdf

    work page 2026

  13. [13]

    On the Height Profile of Analog Error-Correcting Codes,

    R. M. Roth, Z. Zhu, C. Yuan, P. H. Siegel, and A. Jiang, “On the Height Profile of Analog Error-Correcting Codes,” 2026, Online available: https://arxiv.org/abs/2602.20366

    work page arXiv 2026

  14. [14]

    Grassmannian frames with applications to coding and communication,

    T. Strohmer and R. W. Heath, “Grassmannian frames with applications to coding and communication,”Appl. Comput. Harmon. Anal., vol. 14, pp. 257-275 (2003)

    work page 2003

  15. [15]

    Tremain equiangular tight frames,

    M. Fickus, J. Jasper, D. G. Mixon, and J. Peterson, “Tremain equiangular tight frames,”J. Combin. Theory Ser. A, vol. 153, pp. 54-66, Jan. 2018

    work page 2018