pith. sign in

arxiv: 2501.00371 · v4 · submitted 2024-12-31 · 💻 cs.IT · math.IT

Structured Codes for Distributed Matrix Multiplication

Pith reviewed 2026-05-23 06:35 UTC · model grok-4.3

classification 💻 cs.IT math.IT
keywords distributed matrix multiplicationbilinear functionsfinite fieldssum ratesource correlationscompression gainsstructured codes
0
0 comments X

The pith

The sum rate for distributed computation of bilinear functions such as matrix products equals the exact information-theoretic limit when the finite field is large.

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

The paper determines the lowest total communication rate from two nodes holding correlated sources A and B so that a receiver can recover their bilinear combination, including dot products and matrix products over finite fields. It supplies matching upper and lower bounds on this rate that coincide exactly once the field size grows large enough to cover every dimension and many source statistics. A reader would care because the result replaces loose bounds with precise limits and reveals that correlation can yield arbitrarily large rate reductions compared with coding the sources independently. The construction relies on nonlinear maps applied to A and B before linear encoding. This settles the performance limits for an important family of functions that arise in distributed computing.

Core claim

Bounds are established on the optimal sum rate that lets a receiver compute bilinear functions of correlated sources A and B, including dot products and general matrix products over finite fields. The bounds are tight for large field sizes, for which case the exact fundamental performance limits are derived for all problem dimensions and a large class of sources. The achievability uses nonlinear transformations of A and B calibrated to work with linear encoding, while the converses are obtained by calibrating existing bounding methods.

What carries the argument

Nonlinear transformations of the sources A and B that are chosen to interact with linear encoding so the receiver can recover the target bilinear function.

If this is right

  • Exact fundamental performance limits hold for every problem dimension once the field is large.
  • Unbounded compression gains appear relative to separate source coding, and the size of the gain depends on the source correlations.
  • The same rate characterization covers both dot products and general matrix products.
  • The converses remain relatively tight across the considered class of sources.

Where Pith is reading between the lines

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

  • If explicit constructions of the required nonlinear transformations can be found for standard matrix dimensions, the scheme could be implemented directly in finite-field arithmetic.
  • The same bounding strategy may apply to other bilinear or low-degree polynomial functions beyond matrix multiplication.
  • When source correlations are strong, system designers could allocate far less bandwidth to distributed linear algebra tasks than current separate-coding methods require.

Load-bearing premise

Suitable nonlinear transformations of A and B must exist that allow the linear encoding to produce the bilinear function at the claimed rate.

What would settle it

Exhibiting one bilinear function and source pair where the minimal sum rate strictly exceeds the derived upper bound for arbitrarily large field sizes would show the claimed tightness does not hold.

Figures

Figures reproduced from arXiv: 2501.00371 by Derya Malak.

Figure 1
Figure 1. Figure 1: Distributed computation of D = A⊺B. To that end, we take a non-real-time approach that relies on accumulating length n sequences of potentially correlated source matrices. Specifically, distributed sources are block-encoded with blocklength n, aiming to devise (n, ϵ)-coding schemes that approximate the desired matrix prod￾uct with accuracy 1−ϵ, and achieve near-optimal rates, for asymptotically1 lossless c… view at source ↗
Figure 2
Figure 2. Figure 2: Gain, η from Corollary 1. The flat (yellow) line marks η = 1. where U ∈ F m/2×1 q and V ∈ F m/2×1 q are vector variables, and W ∈ Fq is a random variable, and they satisfy the following relations: U = A2 ⊕q B1 , V = A1 ⊕q B2 , W = A ⊺ 2A1 ⊕q B ⊺ 1B2 . (6) Proof. See Appendix A-A. For odd-length vectors, we refer the reader to Appendix A-B. In the scheme of Proposition 1, the receiver, using U, V, and W, ma… view at source ↗
Figure 3
Figure 3. Figure 3: Rate comparisons for various source PMFs. (Left) Corollary 1 for [PITH_FULL_IMAGE:figures/full_fig_p012_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Rate (in log scale) versus p for distributed computing of (Left) symmetric matrices A⊺B = B⊺A via distributed multiplication of matrices A, B ∈ F m×m 2 for different m, and (Right) square matrices via distributed matrix multiplication for different m and l = 2, where the joint source PMF is given in Example 1. Exploiting Proposition 7, to compute D = A⊺B ∈ F 2×2 3 , we can achieve a sum rate of R Σ KM ≤ 2m… view at source ↗
Figure 5
Figure 5. Figure 5: The distributed matrix multiplication framework considered in the current work. The colorings for the input () [PITH_FULL_IMAGE:figures/full_fig_p025_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Communication cost of workers for computing an element of [PITH_FULL_IMAGE:figures/full_fig_p034_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Total communication cost. communication cost from the master is high. Note when mA s ≈ m for MatDot and StMatDot, where s = sr, the master communication costs are identical for both, whereas for StPolyDot the master-communication cost can be made smaller (see Table IV). The total communication costs for PolyDot and StPolyDot can be made equal when mA sr ≈ m sc . In [PITH_FULL_IMAGE:figures/full_fig_p034_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Computation cost per worker for computing [PITH_FULL_IMAGE:figures/full_fig_p035_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Total computation cost. centralized master node configurations in [7], [8] — up to a factor of 2, incurring minimal computation overhead for large memory regimes. StPolyDot codes improve the tradeoff between the communication and computation costs and reveal operating points where structured codes outperform unstructured ones. This flexibility allows designers to refine polynomial coding implementations. T… view at source ↗
Figure 10
Figure 10. Figure 10: (Left) Partitioning of source vectors. (Right) Non-linear transformations of source vectors, given in (97). [PITH_FULL_IMAGE:figures/full_fig_p037_10.png] view at source ↗
read the original abstract

Our work addresses the well-known open problem of distributed computing of bilinear functions of two correlated sources ${\bf A}$ and ${\bf B}$. In a setting with two nodes, with the first node having access to ${\bf A}$ and the second to ${\bf B}$, we establish bounds on the optimal sum rate that allows a receiver to compute an important class of non-linear functions, and in particular bilinear functions, including dot products $\langle {\bf A},{\bf B}\rangle$, and general matrix products ${\bf A}^{\intercal}{\bf B}$ over finite fields. The bounds are tight for large field sizes, for which case we can derive the exact fundamental performance limits for all problem dimensions and a large class of sources. Our achievability scheme involves the design of non-linear transformations of ${\bf A}$ and ${\bf B}$, carefully calibrated to work synergistically with the structured linear encoding scheme by K\"orner and Marton. The subsequent converses derived here, calibrate the Han-Kobayashi approach and the strong converse of Ahlswede-G\'acs-K\"orner to yield relatively tight converses on the sum rate. We exhibit unbounded compression gains over Slepian-Wolf coding, depending on the source correlations. In the end, this work characterizes the fundamental limits of distributed computing for a crucial class of functions, while succinctly capturing the inherent computation structures and source correlations.

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

2 major / 1 minor

Summary. The paper addresses the distributed computation of bilinear functions (including dot products and matrix products A^T B over finite fields) of two correlated sources A and B held at separate nodes. It derives bounds on the minimal sum rate needed for a receiver to recover the function value, claiming these bounds are tight for large field sizes and thereby yield exact fundamental limits for all dimensions and a broad class of sources. Achievability is obtained by designing non-linear transformations of A and B that interact with Körner-Marton structured linear encoding; converses are obtained by calibrating the Han-Kobayashi region and the Ahlswede-Gács-Körner strong converse. The work also exhibits unbounded rate gains relative to Slepian-Wolf coding that depend on source correlation.

Significance. If the claimed tightness holds, the manuscript would characterize the exact rate region for a practically relevant class of distributed bilinear computations, explicitly linking computation structure to source correlation and demonstrating that structured coding can outperform classical source coding by unbounded factors. The calibration of existing converse techniques to this setting is a positive technical contribution.

major comments (2)
  1. [Abstract (achievability scheme description)] The central tightness claim for large field sizes rests on the existence of non-linear transformations of A and B that achieve the Körner-Marton inner bound for the specific bilinear functions (dot product, matrix product). The abstract asserts such transformations are designed, but without an explicit construction, a proof of existence, or a verification that the resulting rates match the converse for the target source statistics, the achievability direction remains unsubstantiated and load-bearing for the exact-limit statement.
  2. [Abstract (converse paragraph)] The converse statements calibrate Han-Kobayashi and Ahlswede-Gács-Körner to the bilinear setting. It is unclear whether the resulting outer bounds are derived under the same non-linear preprocessing or whether they apply directly to the original sources; any mismatch would prevent the claimed tightness even for large fields.
minor comments (1)
  1. [Abstract] Notation for the finite-field matrix product A^T B should be introduced with explicit dimension parameters (m, n, p) at first use to clarify the general case.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the careful reading and constructive feedback. The comments help clarify the presentation of our achievability and converse arguments. We address each major comment below.

read point-by-point responses
  1. Referee: [Abstract (achievability scheme description)] The central tightness claim for large field sizes rests on the existence of non-linear transformations of A and B that achieve the Körner-Marton inner bound for the specific bilinear functions (dot product, matrix product). The abstract asserts such transformations are designed, but without an explicit construction, a proof of existence, or a verification that the resulting rates match the converse for the target source statistics, the achievability direction remains unsubstantiated and load-bearing for the exact-limit statement.

    Authors: The explicit construction of the non-linear transformations, their existence proof via algebraic properties of large finite fields, and the verification that the resulting rates match the converse are all provided in Section III and the proof of Theorem 2. The abstract summarizes this design but does not repeat the full details. To address the concern, we will revise the abstract to briefly indicate that the transformations are constructed in Section III and achieve the Körner-Marton bound for the target statistics. revision: partial

  2. Referee: [Abstract (converse paragraph)] The converse statements calibrate Han-Kobayashi and Ahlswede-Gács-Körner to the bilinear setting. It is unclear whether the resulting outer bounds are derived under the same non-linear preprocessing or whether they apply directly to the original sources; any mismatch would prevent the claimed tightness even for large fields.

    Authors: The outer bounds are derived after the non-linear preprocessing: the Han-Kobayashi region and Ahlswede-Gács-Körner strong converse are calibrated directly to the transformed sources, ensuring the effective correlation structure is accounted for. This is shown in Section IV, where the same preprocessing is used in both directions to obtain tightness for large fields. We will add a clarifying sentence to the abstract and Section IV to make this explicit. revision: partial

Circularity Check

0 steps flagged

No circularity: bounds derived via independent achievability construction and standard converses

full rationale

The derivation relies on a new achievability scheme that designs non-linear transformations of A and B to interact with the existing Körner-Marton linear encoding, paired with converses that calibrate established Han-Kobayashi and Ahlswede-Gács-Körner techniques. These steps introduce independent content (the specific synergistic transformations for bilinear functions) rather than reducing any claimed rate or limit to a fitted parameter, self-definition, or self-citation chain by construction. The tightness result for large field sizes follows from matching the two directions without the equations or limits being tautological to the inputs.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract alone supplies no information on free parameters, background axioms, or newly postulated entities.

pith-pipeline@v0.9.0 · 5766 in / 1070 out tokens · 46585 ms · 2026-05-23T06:35:54.827861+00:00 · methodology

discussion (0)

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

Reference graph

Works this paper leans on

139 extracted references · 139 canonical work pages · 1 internal anchor

  1. [1]

    Distributed structured matrix multiplication,

    D. Malak, “Distributed structured matrix multiplication,” in Proc., IEEE Int. Symp. Inf. Theory (ISIT) , Athens, Greece, Jul. 2024

  2. [2]

    Structured polynomial codes,

    M. R. D. Salehi, A. Tanha, and D. Malak, “Structured polynomial codes,” in Recent Results Poster Session of IEEE ISIT , Athens, Greece, Jul. 2024

  3. [3]

    Structured coded matrix multiplication,

    A. Tanha, M. R. D. Salehi, and D. Malak, “Structured coded matrix multiplication,” submitted, IEEE ISIT , Jan. 2025

  4. [4]

    Strang, Introduction to Linear Algebra

    G. Strang, Introduction to Linear Algebra . Cambridge University Press, 2023

  5. [5]

    Quantum fingerprinting,

    H. Buhrman, R. Cleve, J. Watrous, and R. De Wolf, “Quantum fingerprinting,” Phys. Rev. Lett., vol. 87, no. 16, p. 167902, Sep. 2001

  6. [6]

    Throughput-distortion computation of generic matrix multiplication: Toward a computation channel for digital signal processing systems,

    D. Anastasia and Y . Andreopoulos, “Throughput-distortion computation of generic matrix multiplication: Toward a computation channel for digital signal processing systems,” IEEE Trans. Signal Process. , vol. 60, no. 4, pp. 2024–37, Nov. 2011

  7. [7]

    Polynomial codes: An optimal design for high-dimensional coded matrix multiplication,

    Q. Yu, M. Maddah-Ali, and S. Avestimehr, “Polynomial codes: An optimal design for high-dimensional coded matrix multiplication,” in Proc., Adv. Neural Inf. Process. Syst. , vol. 30, Long Beach, CA, Dec. 2017, pp. 4403–4413

  8. [8]

    On the optimal recovery threshold of coded matrix multiplication,

    S. Dutta, M. Fahim, F. Haddadpour, H. Jeong, V . Cadambe, and P. Grover, “On the optimal recovery threshold of coded matrix multiplication,” IEEE Trans. Inf. Theory , vol. 66, no. 1, pp. 278–301, Jul. 2019. March 14, 2025 DRAFT 76

  9. [9]

    Gradient coding: Avoiding stragglers in distributed learning,

    R. Tandon, Q. Lei, A. G. Dimakis, and N. Karampatziakis, “Gradient coding: Avoiding stragglers in distributed learning,” in Proc., Int. Conf. on Machine Learning , Sydney, Australia, Aug. 2017, pp. 3368–3376

  10. [10]

    Gradient coding from cyclic MDS codes and expander graphs,

    N. Raviv, I. Tamo, R. Tandon, and A. G. Dimakis, “Gradient coding from cyclic MDS codes and expander graphs,” IEEE Trans. Inf. Theory, vol. 66, no. 12, pp. 7475–7489, Dec. 2020

  11. [11]

    Improving distributed gradient descent using Reed-Solomon codes,

    W. Halbawi, N. Azizan, F. Salehi, and B. Hassibi, “Improving distributed gradient descent using Reed-Solomon codes,” in Proc., IEEE ISIT , Vail, CO, Jun. 2018, pp. 2027–2031

  12. [12]

    Analog Lagrange coded computing,

    M. Soleymani, H. Mahdavifar, and A. S. Avestimehr, “Analog Lagrange coded computing,” IEEE J. Sel. Areas Inf. Theory, vol. 2, no. 1, pp. 283–295, Feb. 2021

  13. [13]

    Lagrange coded computing: Optimal design for resiliency, security, and privacy,

    Q. Yu, S. Li, N. Raviv, S. M. M. Kalan, M. Soltanolkotabi, and S. A. Avestimehr, “Lagrange coded computing: Optimal design for resiliency, security, and privacy,” in Proc., Int. Conf. Artif. Intell. Stat. , Naha, Okinawa, Japan, Apr. 2019, pp. 1215–1225

  14. [14]

    Data-intensive cloud computing: Requirements, expectations, challenges, and solutions,

    J. Shamsi, M. A. Khojaye, and M. A. Qasmi, “Data-intensive cloud computing: Requirements, expectations, challenges, and solutions,” J. Grid Comput. , vol. 11, no. 2, pp. 281–310, Jun. 2013

  15. [15]

    Federated learning with lossy distributed source coding: Analysis and optimization,

    H. Yang, T. Ding, and X. Yuan, “Federated learning with lossy distributed source coding: Analysis and optimization,” IEEE Trans. Commun. , vol. 71, no. 8, pp. 4561–4576, May 2023

  16. [16]

    How to encode the modulo-two sum of binary sources (corresp.),

    J. K ¨orner and K. Marton, “How to encode the modulo-two sum of binary sources (corresp.),” IEEE Trans. Inf. Theory , vol. 25, no. 2, pp. 219–221, Mar. 1979

  17. [17]

    A dichotomy of functions F(X, Y) of correlated sources (X, Y),

    T. S. Han and K. Kobayashi, “A dichotomy of functions F(X, Y) of correlated sources (X, Y),” IEEE Trans. Inf. Theory, vol. 33, no. 1, pp. 69–76, Jan. 1987

  18. [18]

    Computation over multiple-access channels,

    B. Nazer and M. Gastpar, “Computation over multiple-access channels,” IEEE Trans. Inf. Theory , vol. 53, no. 10, pp. 3498–3516, Sep. 2007

  19. [19]

    Towards an algebraic network information theory: Distributed lossy computation of linear functions,

    S. H. Lim, C. Feng, A. Pastore, B. Nazer, and M. Gastpar, “Towards an algebraic network information theory: Distributed lossy computation of linear functions,” in Proc., IEEE ISIT , Paris, France, Jun. 2019, pp. 1827–31

  20. [20]

    Linear coding schemes for the distributed computation of subspaces,

    V . Lalitha, N. Prakash, K. Vinodh, P. V . Kumar, and S. S. Pradhan, “Linear coding schemes for the distributed computation of subspaces,” IEEE J. Sel. Areas Commun. , vol. 31, no. 4, pp. 678–690, Mar. 2013

  21. [21]

    Noiseless coding of correlated information sources,

    D. Slepian and J. K. Wolf, “Noiseless coding of correlated information sources,” IEEE Trans. Inf. Theory, vol. 19, no. 4, pp. 471–480, Jul. 1973

  22. [22]

    On source coding with side information via a multiple-access channel and related problems in multi-user information theory,

    R. Ahlswede and T. Han, “On source coding with side information via a multiple-access channel and related problems in multi-user information theory,” IEEE Trans. Inf. Theory , vol. 29, no. 3, pp. 396–412, May 1983

  23. [23]

    Distributed source coding using abelian group codes: A new achievable rate-distortion region,

    D. Krithivasan and S. S. Pradhan, “Distributed source coding using abelian group codes: A new achievable rate-distortion region,” IEEE Trans. Inf. Theory , vol. 57, no. 3, pp. 1495–1519, Feb. 2011

  24. [24]

    An algebraic and probabilistic framework for network information theory,

    S. S. Pradhan, A. Padakandla, and F. Shirani, “An algebraic and probabilistic framework for network information theory,” Found. Trends Commun. Inf. Theory , vol. 18, no. 2, pp. 173–379, Dec. 2020

  25. [25]

    Straggler mitigation in distributed matrix multiplication: Fundamental limits and optimal coding,

    Q. Yu, M. A. Maddah-Ali, and A. S. Avestimehr, “Straggler mitigation in distributed matrix multiplication: Fundamental limits and optimal coding,” IEEE Trans. Inf. Theory , vol. 66, no. 3, pp. 1920–1933, Jan. 2020

  26. [26]

    Secure distributed computing with straggling servers using polynomial codes,

    H. Yang and J. Lee, “Secure distributed computing with straggling servers using polynomial codes,” IEEE Trans. Inf. Foren. and Secur., vol. 14, no. 1, pp. 141–150, Jun. 2018

  27. [27]

    Coded sparse matrix multiplication,

    S. Wang, J. Liu, and N. Shroff, “Coded sparse matrix multiplication,” in Proc., Int. Conf. Mach. Learn. , Jul. 2018, pp. 5152–5160

  28. [28]

    Distributed matrix-vector multiplication: A convolutional coding approach,

    A. B. Das and A. Ramamoorthy, “Distributed matrix-vector multiplication: A convolutional coding approach,” in Proc., IEEE ISIT, Paris, France, Jul. 2019, pp. 3022–3026

  29. [29]

    Distributed matrix multiplication with straggler tolerance using algebraic function fields,

    A. Fidalgo-D ´ıaz and U. Mart ´ınez-Pe˜nas, “Distributed matrix multiplication with straggler tolerance using algebraic function fields,” arXiv preprint arXiv:2401.13573 , Jan. 2024

  30. [30]

    Numerically stable polynomially coded computing,

    M. Fahim and V . R. Cadambe, “Numerically stable polynomially coded computing,” IEEE Trans. Inf. Theory , vol. 67, no. 5, pp. 2758–2785, Jan. 2021

  31. [31]

    Distributed matrix computations with low-weight encodings,

    A. B. Das, A. Ramamoorthy, D. J. Love, and C. G. Brinton, “Distributed matrix computations with low-weight encodings,” IEEE J. Sel. Areas Inf. Theory , Aug. 2023

  32. [32]

    Random Khatri-Rao-product codes for numerically-stable distributed matrix multiplication,

    A. M. Subramaniam, A. Heidarzadeh, and K. R. Narayanan, “Random Khatri-Rao-product codes for numerically-stable distributed matrix multiplication,” in Proc., Annu. Allerton Conf. Commun. Control Comput. (Allerton) , Monticello, IL, Sep. 2019, pp. 253–259

  33. [33]

    On the capacity of secure distributed matrix multiplication,

    W.-T. Chang and R. Tandon, “On the capacity of secure distributed matrix multiplication,” in Proc., IEEE Global Commun. Conf. (Globecom), Abu Dhabi, UAE, Dec. 2018, pp. 1–6

  34. [34]

    On the capacity of secure distributed batch matrix multiplication,

    Z. Jia and S. A. Jafar, “On the capacity of secure distributed batch matrix multiplication,” IEEE Trans. Inf. Theory , vol. 67, no. 11, pp. 7420–7437, Sep. 2021

  35. [35]

    GASP codes for secure distributed matrix multiplication,

    R. G. L. D’Oliveira, S. El Rouayheb, and D. Karpuk, “GASP codes for secure distributed matrix multiplication,” IEEE Trans. Inf. Theory, vol. 66, no. 7, pp. 4038–4050, Feb. 2020

  36. [36]

    Degree tables for secure distributed matrix multiplication,

    R. G. L. D’Oliveira, S. El Rouayheb, D. Heinlein, and D. Karpuk, “Degree tables for secure distributed matrix multiplication,” IEEE J. Sel. Areas Inf. Theory , vol. 2, no. 3, pp. 907–918, Aug. 2021

  37. [37]

    Preserving sparsity and privacy in straggler-resilient distributed matrix computations,

    A. B. Das, A. Ramamoorthy, D. J. Love, and C. G. Brinton, “Preserving sparsity and privacy in straggler-resilient distributed matrix computations,” in Proc., IEEE Allerton , Monticello, IL, Sep. 2023, pp. 1–8

  38. [38]

    Secure MatDot codes: A secure, distributed matrix multiplication scheme,

    H. H. L ´opez, G. L. Matthews, and D. Valvo, “Secure MatDot codes: A secure, distributed matrix multiplication scheme,” in Proc., IEEE ITW , Mumbai, India, Nov. 2022, pp. 149–154

  39. [39]

    Improved constructions for secure multi-party batch matrix multiplication,

    J. Zhu, Q. Yan, and X. Tang, “Improved constructions for secure multi-party batch matrix multiplication,” IEEE Trans. Commun., vol. 69, no. 11, pp. 7673–7690, Aug. 2021

  40. [40]

    Private and secure distributed matrix multiplication with flexible communi- cation load,

    M. Aliasgari, O. Simeone, and J. Kliewer, “Private and secure distributed matrix multiplication with flexible communi- cation load,” IEEE Trans. Inf. Forensics Secur., vol. 15, pp. 2722–2734, Feb. 2020

  41. [41]

    Computing linear transformations with unreliable components,

    Y . Yang, P. Grover, and S. Kar, “Computing linear transformations with unreliable components,” IEEE Trans. Inf. Theory, vol. 63, no. 6, pp. 3729–3756, Apr. 2017

  42. [42]

    Masterless coded computing: A fully-distributed coded FFT algorithm,

    H. Jeong, T. M. Low, and P. Grover, “Masterless coded computing: A fully-distributed coded FFT algorithm,” in Proc., Allerton, Monticello, IL, Oct. 2018, pp. 887–894

  43. [43]

    A unified coded deep neural network training strategy based on generalized PolyDot codes,

    S. Dutta, Z. Bai, H. Jeong, T. M. Low, and P. Grover, “A unified coded deep neural network training strategy based on generalized PolyDot codes,” in Proc., IEEE ISIT , Vail, CO, Jun. 2018, pp. 1585–1589

  44. [44]

    Can a noisy encoder be used to communicate reliably?

    Y . Yang, P. Grover, and S. Kar, “Can a noisy encoder be used to communicate reliably?” in Proc., Allerton, Monticello, IL, Sep. 2014, pp. 659–666. DRAFT March 14, 2025 77

  45. [45]

    Reliable information storage in memories designed from unreliable components,

    M. G. Taylor, “Reliable information storage in memories designed from unreliable components,” Bell System Technical Journal, vol. 47, no. 10, pp. 2299–2337, Dec. 1968

  46. [46]

    Wyner-Ziv theory for a general function of the correlated sources,

    H. Yamamoto, “Wyner-Ziv theory for a general function of the correlated sources,” IEEE Trans. Inf. Theory , vol. 28, no. 5, pp. 803–7, Sep. 1982

  47. [47]

    Coding of an information source having ambiguous alphabet and the entropy of graphs,

    J. K ¨orner, “Coding of an information source having ambiguous alphabet and the entropy of graphs,” in Proc., 6th Prague Conf. Inf. Theory , Prague, Czech Republic, Sep. 1973, pp. 411–425

  48. [48]

    Coding for computing,

    A. Orlitsky and J. R. Roche, “Coding for computing,” IEEE Trans. Inf. Theory , vol. 47, no. 3, p. 903–917, Mar. 2001

  49. [49]

    On network functional compression,

    S. Feizi and M. M ´edard, “On network functional compression,” IEEE Trans. Inf. Theory , vol. 60, no. 9, pp. 5387–5401, Sep. 2014

  50. [50]

    On Computing a Function of Correlated Sources

    M. Sefidgaran and A. Tchamkerten, “On computing a function of correlated sources,” arXiv preprint arXiv:1107.5806 , Jul. 2011

  51. [51]

    Fractional graph coloring for functional compression with side information,

    D. Malak, “Fractional graph coloring for functional compression with side information,” in Proc., IEEE ITW , Mumbai, India, Nov. 2022

  52. [52]

    Weighted graph coloring for quantized computing,

    ——, “Weighted graph coloring for quantized computing,” in Proc., IEEE ISIT, Taipei, Taiwan, Jun. 2023, pp. 2290–2295

  53. [53]

    An achievable low complexity encoding scheme for coloring cyclic graphs,

    M. R. D. Salehi and D. Malak, “An achievable low complexity encoding scheme for coloring cyclic graphs,” in Proc., Allerton, Monticello, IL, Sep. 2023, pp. 1–8

  54. [54]

    Multi-server multi-function distributed computation,

    D. Malak, M. R. Deylam Salehi, B. Serbetci, and P. Elia, “Multi-server multi-function distributed computation,” Entropy, vol. 26, no. 6, p. 448, Jun. 2024

  55. [55]

    Multi-functional distributed computing,

    ——, “Multi-functional distributed computing,” in Proc., IEEE Allerton , Urbana-Champaign, IL, 2024, pp. 1–8

  56. [56]

    Function-correcting codes,

    A. Lenz, R. Bitar, A. Wachter-Zeh, and E. Yaakobi, “Function-correcting codes,” IEEE Trans. Inf. Theory , May 2023

  57. [57]

    On K ¨orner-Marton’s sum modulo two problem,

    M. Sefidgaran, A. Gohari, and M. R. Aref, “On K ¨orner-Marton’s sum modulo two problem,” inProc., Iran Wksh. Commun. and Inf. Theory , May 2015, pp. 1–6

  58. [58]

    On optimal weighted-sum rates for the modulo sum problem,

    C. Nair and Y . N. Wang, “On optimal weighted-sum rates for the modulo sum problem,” in Proc., IEEE ISIT, Jun. 2020, pp. 2416–2420

  59. [59]

    Expand-and-randomize: An algebraic approach to secure computation,

    Y . Zhao and H. Sun, “Expand-and-randomize: An algebraic approach to secure computation,” Entropy, vol. 23, no. 11, p. 1461, Nov. 2021

  60. [60]

    How to securely compute the modulo-two sum of binary sources,

    D. Data, B. K. Dey, M. Mishra, and V . M. Prabhakaran, “How to securely compute the modulo-two sum of binary sources,” in Proc., IEEE ITW , Hobart, Tasmania, Australia, Nov. 2014, pp. 496–500

  61. [61]

    To get a bit of information may be as hard as to get full information,

    R. Ahlswede and I. Csisz ´ar, “To get a bit of information may be as hard as to get full information,” IEEE Trans. Inf. Theory, vol. 27, no. 4, pp. 398–408, Jul. 1981

  62. [62]

    Unified approach for computing sum of sources over CQ-MAC,

    M. A. Sohail, T. A. Atif, and S. S. Pradhan, “Unified approach for computing sum of sources over CQ-MAC,” in Proc., IEEE ISIT, Espoo, Finland, 2022, pp. 1868–1873

  63. [63]

    Abelian group codes for channel coding and source coding,

    A. G. Sahebi and S. S. Pradhan, “Abelian group codes for channel coding and source coding,” IEEE Trans. Inf. Theory , vol. 61, no. 5, pp. 2399–2414, Feb. 2015

  64. [64]

    Corrections to “Abelian group codes for channel coding and source coding

    S. S. Pradhan, M. Heidari, and A. G. Sahebi, “Corrections to “Abelian group codes for channel coding and source coding”[May 15 2399-2414],” IEEE Trans. Inf. Theory , vol. 64, no. 5, pp. 3953–3953, Jan. 2018

  65. [65]

    On distributed source coding using Abelian group codes,

    A. G. Sahebi and S. S. Pradhan, “On distributed source coding using Abelian group codes,” in Proc., Allerton, Monticello, IL, Oct. 2012, pp. 2068–2074

  66. [66]

    On the capacity of abelian group codes over discrete memoryless channels,

    ——, “On the capacity of abelian group codes over discrete memoryless channels,” in Proc., IEEE ISIT , Jul. 2011, pp. 1743–1747

  67. [67]

    Distributed computing of functions of structured sources with helper side information,

    D. Malak, “Distributed computing of functions of structured sources with helper side information,” in Proc., IEEE Int. Wksh. Signal Proces. Advances in Wireless Commun. (SPAWC) , Shanghai, China, Sep. 2023

  68. [68]

    Using spatial principles to optimize distributed computing for enabling the physical science discoveries,

    C. Yang, H. Wu, Q. Huang, Z. Li, and J. Li, “Using spatial principles to optimize distributed computing for enabling the physical science discoveries,” Proc., Natl. Acad. Sci. U.S.A. , vol. 108, no. 14, pp. 5498–5503, Apr. 2011

  69. [69]

    An overview of the bioextract server: A distributed, web-based system for genomic analysis,

    C. Lushbough and V . Brendel, “An overview of the bioextract server: A distributed, web-based system for genomic analysis,” in Proc., Adv. Comput. Biol. New York, NY: Springer, Jan. 2010, pp. 361–369

  70. [70]

    MapReduce: Simplified data processing on large clusters,

    J. Dean and S. Ghemawat, “MapReduce: Simplified data processing on large clusters,” Commun. ACM, vol. 51, no. 1, pp. 107–113, Jan. 2008

  71. [71]

    John Wiley & Sons, 2014

    EMC Education Services, Data Science and Big Data Analytics: Discovering, Analyzing, Visualizing and Presenting Data. John Wiley & Sons, 2014

  72. [72]

    Spark: Cluster computing with working sets,

    M. Zaharia, M. Chowdhury, M. J. Franklin, S. Shenker, and I. Stoica, “Spark: Cluster computing with working sets,” in Proc., USENIX HotTop Cloud Comp. Works. , Boston, MA, USA, Jun. 2010

  73. [73]

    Communication-efficient distributed monitoring of thresholded counts,

    R. Keralapura, G. Cormode, and J. Ramamirtham, “Communication-efficient distributed monitoring of thresholded counts,” in Proc., ACM SIGMOD Int. Conf. Management of Data , New York, NY , USA, Jun. 2006, p. 289–300

  74. [74]

    Flexible constructions for distributed matrix multiplication,

    W. Li, Z. Chen, Z. Wang, S. A. Jafar, and H. Jafarkhani, “Flexible constructions for distributed matrix multiplication,” in Proc., IEEE ISIT , Virtual Conference, Jul. 2021, pp. 1576–1581

  75. [75]

    Distributed resource allocation and computation offloading in fog and cloud networks with non-orthogonal multiple access,

    Y . Liu, F. R. Yu, X. Li, H. Ji, and V . C. Leung, “Distributed resource allocation and computation offloading in fog and cloud networks with non-orthogonal multiple access,” IEEE Trans. Veh. Tech., vol. 67, no. 12, pp. 12 137–51, Sep. 2018

  76. [76]

    Datacenter traffic control: Understanding techniques and tradeoffs,

    M. Noormohammadpour and C. S. Raghavendra, “Datacenter traffic control: Understanding techniques and tradeoffs,” IEEE Commun. Surv. Tutor., vol. 20, no. 2, pp. 1492–1525, Dec. 2017

  77. [77]

    Load distributing for locally distributed systems,

    N. Shivaratri, P. Krueger, and M. Singhal, “Load distributing for locally distributed systems,” Computer, vol. 25, no. 12, pp. 33–44, Dec. 1992

  78. [78]

    Demand-based document dissemination to reduce traffic and balance load in distributed information systems,

    A. Bestavros, “Demand-based document dissemination to reduce traffic and balance load in distributed information systems,” in Proc., IEEE Symp. Parallel Distrib. Process. , San Antonio, Texas, USA, Oct. 1995, pp. 338–345

  79. [79]

    Distributed linearly separable computation,

    K. Wan, H. Sun, M. Ji, and G. Caire, “Distributed linearly separable computation,” IEEE Trans. Inf. Theory , vol. 68, no. 2, pp. 1259–1278, Nov. 2021

  80. [80]

    Multi-user linearly-separable distributed computing,

    A. Khalesi and P. Elia, “Multi-user linearly-separable distributed computing,” IEEE Trans. Inf. Theory , vol. 69, no. 10, pp. 6314–39, Jun. 2023

Showing first 80 references.