Pith. sign in

REVIEW 4 major objections 4 minor 43 references

Relaxation-Free Min-k-Partition for PCI Assignment in 5G Networks

T0 review · 4 major / 4 minor · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read This paper claims that 5G cell identity assignment can be decomposed into a sequence of partition and coloring subproblems, solved by a relaxation-free penalized mirror descent, cutting runtime by up to 20 times while eliminating…

desk verdict A genuinely useful CRT-based decomposition of PCI assignment with a borrowed but well-packaged solver, undermined by overstated claims and a lexicographic step that lacks an optimality guarantee. read the letter →

arxiv 2506.10362 v2 pith:GOIG5IVQ submitted 2025-06-12 eess.SP

Yeqing Qiu , Chengpiao Huang , Ye Xue , Zhipeng Jiang , Qingjiang Shi , Dong Zhang , Zhi-Quan Luo This is my paper · ORCID
classification eess.SP MSC 90C2705C1590C30
keywords PCIassignmentMin-k-PartitionChineseremaindertheoremPenalizedmirrordescentGraphcoloringmod-3interferencemod-305Gnetworkoptimization
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper claims that 5G physical cell identity (PCI) assignment, a hard combinatorial problem, can be split into a fixed sequence of smaller tasks: partition cells into 3 groups by mod-3 value, then into 10 groups by mod-10 value, and finally color the quotient graph to prevent collisions and confusions. The Chinese Remainder Theorem makes the split valid, because a mod-30 value is uniquely determined by the mod-3 and mod-10 pair. To solve each partitioning step, the paper introduces a relaxation-free reformulation of Min-k-Partition as a quadratic program with a norm-equality constraint, solved by a penalized mirror descent algorithm with convergence guarantees. On real urban datasets with up to 10,000 cells, the method is reported to eliminate collisions and confusions while cutting runtime up to 20 times versus prior optimization baselines. If correct, this would make real-time PCI re-optimization practical in large live networks.

What carries the argument

The load-bearing object is the norm-equality characterization of one-hot vectors: for $q > p \ge 1$, a nonzero vector has exactly one nonzero entry if and only if $\|x\|_p = \|x\|_q$ (Lemma 1). With $p=1$, $q=2$, this turns the discrete Min-$k$-Partition constraint $x_i \in \{e_1,\dots,e_k\}$ into a smooth equality constraint on the probability simplex, giving a quadratic program. The paper penalizes the violation $\|x_i\|_1^2 - \|x_i\|_2^2$, which simplifies to the objective $\operatorname{Tr}(X(W - \tfrac{\rho}{2} I)X^T)$ plus a constant, and Theorem 1 proves exactness when $\rho > 2\lambda_{\max}(W)$. The mirror descent update has a closed form using the KL divergence, so each inner iteration is a cheap row-wise softmax-like step, and the convergence rate in Theorem 2 is $O(\sqrt{1/\epsilon_1})$ in the KL stopping criterion.

What would settle it

On small networks where exhaustive search is feasible, compare the decomposition's outcome against the true Pareto frontier of the original problem (P1); any instance where the decomposition's (mod-3, mod-30) interference pair is strictly dominated by some feasible assignment would falsify the claim that the sequential split preserves optimality.

Watch

Extended reading notes

Core claim

The paper's central claim is that the four-way PCI assignment objective, minimizing mod-3 interference, minimizing mod-30 interference, and eliminating collisions and confusions, can be decoupled into stages, each of which is a tractable subproblem. Writing each PCI as $30q_i + r_i$, the split first fixes the mod-3 value of each cell, then fixes the mod-10 value inside each mod-3 class, and finally assigns quotients $q_i$ by greedy graph coloring on 30 separate clusters. For the NP-hard Min-$k$-Partition subproblems, the paper reformulates them as minimizing $\sum_{i,j} [W]_{i,j} x_i^T x_j$ over columns $x_i$ in the probability simplex with the equality constraint $\|x_i\|_p = \|x_i\|_q$, which forces each column to be a one-hot vector; a penalized mirror descent drives the columns to vertices and a local search cleans up the result. On real-world dense urban data, the method is reported to eliminate all collisions and confusions, reduce mod-3 interference by 8.3% and mod-30 interference by 99.7% relative to the strongest baselines, and run up to 20 times faster.

Load-bearing premise

The sequential decomposition assumes that first fixing mod-3 values, then mod-10 values, then quotients loses nothing important compared with jointly optimizing all four objectives, and the paper gives no optimality or approximation guarantee for that split.

Editorial extensions

If this is right

  • Operators can re-optimize only a chosen subset of cells, so a live network can be improved without re-assigning every PCI and destabilizing handovers.
  • Because the coloring stage splits into 30 independent cluster problems, the method is naturally parallel and its runtime is dominated by the largest mod-30 cluster, not the full network size.
  • The same CRT-style decomposition applies to any single modulus and to any set of coprime moduli, making the framework a template for other modular-interference problems.
  • If the reported real-data results hold, PCI re-optimization time drops from weeks, where the binary quadratic programming baseline times out, to minutes or hours, enabling more frequent re-planning.
  • The relaxation-free Min-$k$-Partition reformulation is a general graph-partitioning tool, applicable outside PCI assignment wherever a symmetric weighted graph must be split into $k$ parts.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The decomposition's exactness is asserted for the reformulation, not for the sequential split: nothing in the paper bounds the gap between the two-stage mod-3 and mod-10 solution and the true Pareto frontier of the original problem, so a small-instance exhaustive comparison would be the direct test.
  • The relaxation-free label applies to the reformulation and the penalty equivalence, not to the optimization pathway: mirror descent on a nonconvex problem converges to a stationary point, so in adversarial instances the returned solution may be locally optimal rather than globally optimal.
  • The quotient stage is feasible only when each mod-30 cluster can be colored with at most 34 quotients, since the PCI range 0 to 1007 permits quotients 0 to 33; for denser clusters the paper's own post-processing trades range violations for possible collisions and confusions, which is the regime where a joint approach would have the clearest advantage.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 4 minor

Summary. The paper proposes a new framework for 5G Physical Cell Identity (PCI) assignment. It decomposes the problem, via the Chinese Remainder Theorem, into sequential Min-3-Partition, Min-10-Partition, and graph coloring subproblems. For the Min-k-Partition component, it introduces an exact reformulation as a quadratic program with a norm-equality constraint (P4) and a penalized mirror descent (PMD) algorithm, with Theorem 1 establishing exactness of the penalized problem. The method is evaluated on synthetic random geometric graphs and real-world data from a dense urban network with up to 10,000 cells, reporting near-zero mod-30 interference, zero collisions/confusions for GP-PMD, and significant runtime reductions over several baselines.

Significance. If the claims hold, the paper offers a scalable and practical approach to PCI assignment, a problem of clear industrial relevance. The exact penalized reformulation in Theorem 1 is a sound theoretical contribution and is applicable to general Min-k-Partition beyond PCI. The use of the CRT to decouple modular constraints is a creative and potentially useful idea. The paper also includes an unusually thorough experimental evaluation on real-world data. However, the central sequential decomposition lacks an optimality guarantee, and several headline claims are not supported by the presented evidence, so the significance is presently conditional on additional analysis or validation.

major comments (4)
  1. [Section III-B] The decomposition into sequential Min-3-Partition and Min-10-Partition is lexicographic and is not justified for the multi-objective problem (P1). The text states 'we can break down the problem into first optimizing over r to minimize both mod-3 interference and mod-30 interference,' but this is not what the two-stage procedure does: the mod-30 objective is minimized only within the mod-3 classes fixed by the first stage. A different mod-3 assignment with slightly higher mod-3 cost can dramatically reduce mod-30 cost, so the returned solution can be far from the Pareto frontier of (P1). No approximation bound or Pareto optimality analysis is provided. I recommend adding an experimental comparison with a joint scalarized objective (e.g., a weighted sum) on small instances, or a theoretical bound on the lexicographic gap, to validate the decomposition.
  2. [Section IV-B3 (Theorem 2)] Theorem 2 claims that the mirror descent sequence converges to a stationary point of P5 at a rate T ≥ sqrt((2||W||_{1,1}+ρN)/ϵ1) with respect to the stopping criterion KL(X^{(m+1)}, X^{(m)}) ≤ ϵ1. However, the proof in Appendix D, specifically Eq. (33)–(34), only establishes an upper bound on min_{m=0,...,T-1} KL(X^{(m+1)}, X^{(m)}). This does not imply that the final iterate is a stationary point, nor that the iterates converge. The stated O(1/sqrt(ϵ)) rate applies to the minimum over all iterates, not to the last iterate. Please restate the theorem to match what is actually proven, or provide a stronger convergence result for the final iterate.
  3. [Section III-C and Section VI] The paper repeatedly claims that the framework 'eliminates collisions and confusions,' but Section III-C explicitly acknowledges that the greedy graph coloring may require a post-processing procedure that re-assigns quotients 'at the cost of potentially increasing the numbers of collisions and confusions.' No sufficient conditions are given under which collision-free and confusion-free assignments are guaranteed. The experimental results show zero collisions and confusions for GP-PMD on the tested instances, but this is an empirical observation. The claim should be qualified to 'the method aims to eliminate' or a condition for guaranteed elimination should be stated.
  4. [Section VI-C (Tables II–V)] The headline performance claims are not tied to a consistent baseline. For N=10000, GP-PMD's runtime of 6185s is slower than GGC (254s), and the reported 'up to 20 times' speedup appears to be relative only to the Genetic method (55515s). Similarly, the claimed 8.3% mod-3 reduction and 99.7% mod-30 reduction do not state against which method they are computed; the reductions vary substantially across baselines. Please report relative improvements with explicit baseline names, or compare to the best-performing baseline for each metric, so that the claims are reproducible.
minor comments (4)
  1. [Abstract] The abstract uses 'Min-k-Partition' generically, but the specific cases k=3 and k=10 are central to the application; please be explicit about these values where they matter.
  2. [Section IV-E] The outer-loop termination criterion in Algorithm 2 is based on the off-diagonal orthogonality of QX^τ. The paper does not discuss how this criterion relates to the exactness threshold ρ0 of Theorem 1, leaving the practical choice of ρ unconnected to the theoretical guarantee.
  3. [Section V] The notation ``er_i'' is introduced in Section V-A without explicit definition; it should be defined as the mod-k value of cell i for clarity.
  4. [Supplementary Material] Reference [25] is a self-citation to the arXiv version of the same paper and is used as supplementary material. In an anonymous review setting, this should be replaced by an anonymized supplement.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the reformulation and exactness proofs are self-contained, and the lexicographic decomposition is a heuristic gap rather than a circular step.

full rationale

The paper's derivation chain does not reduce any claimed prediction or first-principles result to its own inputs. The change of variables PCI = 30q + r in Section III-A is an exact reformulation of (P1) as (P2), and the split into a mod-30 value assignment followed by a quotient assignment is exact for the stated objectives because both objective terms depend only on r. The further split in Section III-B into a mod-3 assignment (S1.1) and then a mod-10 assignment (S1.2.r) is an explicit sequential, lexicographic heuristic; the paper does not prove this split is equivalent to the joint minimization of both objectives, so any optimality gap is a correctness or approximation concern, not circularity. The Min-k-Partition reformulation (P4) relies on Lemma 1, which is stated with a proof in Appendix A, and the exactness of the penalized problem (P5) is proved in Appendix C via strict concavity and extreme-point arguments; these are self-contained mathematical proofs rather than fitted assumptions. Theorem 2's convergence argument is also proved in Appendix D. The empirical claims are benchmark evaluations against external baselines, and the algorithm hyperparameters are adopted from the prior work [32] rather than fitted to the target data. The only self-reference is [25], which is the paper's own arXiv posting used to point to supplementary algorithms and experimental details; it is not load-bearing evidence for any theorem or empirical result and does not make any derived quantity equal to its input. No circular step can be exhibited from the paper's own equations or citations.

Assumptions & free parameters 4 free parameters · 3 assumptions · 0 invented entities

The central claim rests on the exactness of the norm-equality reformulation (Theorem 1), which is sound, but also on the unproven lexicographic decomposition and on hyperparameters inherited from [32]. No new physical entities are introduced.

free parameters (4)
  • rho0 = 1e-8
    Initial penalty parameter in PMD; chosen from [32] and used in all experiments (Section VI-B-1).
  • gamma = 1.1
    Penalty increase factor per outer iteration; from [32].
  • epsilon1 = 1e-5
    Inner mirror descent stopping tolerance; from [32].
  • epsilon2 = 1e-10
    Outer loop orthogonality termination tolerance; from [32].
assumptions (3)
  • ad hoc to paper Sequential lexicographic minimization of mod-3 then mod-10 interference yields a jointly optimal or near-optimal solution to the original two-objective PCI problem (P1).
    Section III-B states the decomposition without proof; no Pareto optimality or approximation guarantee is provided. If false, the CRT-based method may trade mod-30 performance for mod-3 performance suboptimally.
  • domain assumption The interference matrix W, first-order neighbor set E1, and second-order neighbor set E2 are accurately estimated and known.
    Section II-A: 'we assume that W, E1, and E2 have been estimated and are known.' This is standard for PCI planning but unverifiable without the proprietary data.
  • domain assumption A valid quotient assignment exists within the PCI range 0 to 1007 for the given graph clusters, and the greedy re-assignment heuristic can repair range violations without creating excessive collisions or confusions.
    Section III-C introduces a heuristic post-processing because 'The graph coloring method does not always guarantee that the PCI range constraint ... is satisfied.' The claim of eliminating collisions and confusions depends on this repair succeeding.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Relaxation-Free Min-k-Partition for PCI Assignment in 5G Networks." pith.science (2026). https://pith.science/paper/GOIG5IVQ

@misc{pith2026250610362,
  author       = {Pith},
  title        = {Pith review of: Relaxation-Free Min-k-Partition for PCI Assignment in 5G Networks},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/GOIG5IVQ}},
  note         = {Machine review of arXiv:2506.10362}
}
read the original abstract

Physical Cell Identity (PCI) is a critical parameter in 5G networks. Efficient and accurate PCI assignment is essential for mitigating mod-3 interference, mod-30 interference, collisions, and confusions among cells, which directly affect network reliability and user experience. In this paper, we propose a novel framework for PCI assignment by decomposing the problem into Min-3-Partition, Min-10-Partition, and a graph coloring problem, leveraging the Chinese Remainder Theorem (CRT). Furthermore, we develop a relaxation-free approach to the general Min-k-Partition problem by reformulating it as a quadratic program with a norm-equality constraint and solving it using a penalized mirror descent (PMD) algorithm. The proposed method demonstrates superior computational efficiency and scalability, significantly reducing interference while eliminating collisions and confusions in large-scale 5G networks. Numerical evaluations on real-world datasets show that our approach reduces computational time by up to 20 times compared to state-of-the-art methods, making it highly practical for real-time PCI optimization in large-scale networks. These results highlight the potential of our method to improve network performance and reduce deployment costs in modern 5G systems.

Figures

Figures reproduced from arXiv: 2506.10362 by the authors.

Figure 1
Figure 1. Illustration of a 5G wireless network comprising [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Illustration of the proposed PCI assignment framework. [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Convergence curve for mod-3 optimization problem of MD with KL divergence. constant L accounts for the curvature of the objective in early exploration, and the O(1/m) decay mitigates oscillation in later stages. Moreover, as the ℓ1,1-norm of matrix W, the penalized parameter ρ, and the number of cells N increase while keeping ϵ1 fixed, the number of iterations required for convergence will also increase. Upon obtain… view at source ↗
Figures from the paper (11 more)
Figure 4
Figure 4. Figure 4: Experiment results on random geometric graphs. [PITH_FULL_IMAGE:figures/full_fig_p011_4.png]
Figure 5
Figure 5. Figure 5: Boxplot of collision across methods and graph types. [PITH_FULL_IMAGE:figures/full_fig_p018_5.png]
Figure 6
Figure 6. Figure 6: Boxplot of confusion across methods and graph types. [PITH_FULL_IMAGE:figures/full_fig_p018_6.png]
Figure 7
Figure 7. Figure 7: Boxplot of mod-3 interference across methods and graph types. [PITH_FULL_IMAGE:figures/full_fig_p019_7.png]
Figure 8
Figure 8. Figure 8: Boxplot of mod-30 interference across methods and graph types. [PITH_FULL_IMAGE:figures/full_fig_p019_8.png]
Figure 9
Figure 9. Figure 9: Boxplot of computational time across methods and graph types. [PITH_FULL_IMAGE:figures/full_fig_p020_9.png]
Figure 10
Figure 10. Figure 10: GP-PMD versus other algorithms on number of collisions [PITH_FULL_IMAGE:figures/full_fig_p021_10.png]
Figure 11
Figure 11. Figure 11: GP-PMD versus other algorithms on number of confusions [PITH_FULL_IMAGE:figures/full_fig_p021_11.png]
Figure 12
Figure 12. Figure 12: GP-PMD versus other algorithms on mod-3 interference [PITH_FULL_IMAGE:figures/full_fig_p022_12.png]
Figure 13
Figure 13. Figure 13: GP-PMD versus other algorithms on mod-30 interference [PITH_FULL_IMAGE:figures/full_fig_p022_13.png]
Figure 14
Figure 14. Figure 14: GP-PMD versus other algorithms on computational time [PITH_FULL_IMAGE:figures/full_fig_p023_14.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

43 extracted references · 35 canonical work pages

  1. [1]

    Signal processing for the 5G revolution,

    R. W. Heath, G. Laus, T. Q. Quek, S. Talwar, and P. Zhou, “Signal processing for the 5G revolution,” IEEE Signal Process. Mag. , vol. 31, no. 6, pp. 12–13, 2014, doi:10.1109/MSP.2014.2345430

    arXiv 2014

  2. [2]

    5G: A tutorial overview of standards, tri- als, challenges, deployment, and practice,

    M. Shafi, A. F. Molisch, P. J. Smith, T. Haustein, P. Zhu, P. De Silva, F. Tufvesson, A. Benjebbour, and G. Wunder, “5G: A tutorial overview of standards, tri- als, challenges, deployment, and practice,” IEEE J. Sel. Areas Commun. , vol. 35, no. 6, pp. 1201–1221, 2017, doi:10.1109/JSAC.2017.2692307

    work page arXiv 2017

  3. [3]

    A survey of 5G network: Architecture and emerging technologies,

    A. Gupta and R. K. Jha, “A survey of 5G network: Architecture and emerging technologies,” IEEE Access , vol. 3, pp. 1206–1232, 2015, doi:10.1109/ACCESS.2015.2461602

    work page arXiv 2015

  4. [4]

    5G deployment: Standalone vs. non- standalone from the operator perspective,

    G. Liu, Y . Huang, Z. Chen, L. Liu, Q. Wang, and N. Li, “5G deployment: Standalone vs. non- standalone from the operator perspective,” IEEE Com- mun. Mag. , vol. 58, no. 11, pp. 83–89, 2020, doi:10.1109/MCOM.001.2000230

    work page doi:10.1109/mcom.001.2000230 2020

  5. [5]

    The quality of experience perspective toward 5g technology,

    L. Pierucci, “The quality of experience perspective toward 5g technology,” IEEE Wireless Commun., vol. 22, no. 4, pp. 10–16, 2015, doi:10.1109/MWC.2015.7224722

    work page arXiv 2015

  6. [6]

    SRCON: A data-driven network perfor- mance simulator for real-world wireless networks,

    Z.-Q. Luo, X. Zheng, D. López-Pérez, Q. Yan, X. Chen, N. Wang, Q. Shi, T.-H. Chang, and A. Garcia- Rodriguez, “SRCON: A data-driven network perfor- mance simulator for real-world wireless networks,” IEEE Commun. Mag. , vol. 61, no. 6, pp. 96–102, 2023, doi:10.1109/MCOM.001.2200179

    work page doi:10.1109/mcom.001.2200179 2023

  7. [7]

    A survey of recent advances in optimization methods for wireless communications,

    Y .-F. Liu, T.-H. Chang, M. Hong, Z. Wu, A. Man-Cho So, E. A. Jorswieck, and W. Yu, “A survey of recent advances in optimization methods for wireless communications,” IEEE J. Sel. Areas Commun. , vol. 42, no. 11, pp. 2992– 3031, 2024, doi:10.1109/JSAC.2024.3443759

    work page arXiv 2024

  8. [8]

    PCI planning based on binary quadratic programming in LTE/LTE-A net- works,

    J. Gui, Z. Jiang, and S. Gao, “PCI planning based on binary quadratic programming in LTE/LTE-A net- works,” IEEE Access , vol. 7, pp. 203–214, 2019, doi:10.1109/ACCESS.2018.2885313

    work page arXiv 2019

Show all 43 references
  1. [9]

    A programmable distributed optimization framework for 5G,

    S. Seetharaman and D. Krishnaswamy, “A programmable distributed optimization framework for 5G,” CSI Trans. ICT, vol. 8, no. 1, pp. 65–68, 2020, doi:10.1007/s40012- 020-00271-x

    2020 doi

  2. [10]

    Interference self-coordination: A proposal to enhance reliability of system-level information in OFDM-based mobile networks via PCI planning,

    H. Yang, A. Huang, R. Gao, T. Chang, and L. Xie, “Interference self-coordination: A proposal to enhance reliability of system-level information in OFDM-based mobile networks via PCI planning,” IEEE Trans. Wire- less Commun. , vol. 13, no. 4, pp. 1874–1887, 2014, doi:10.1109/TW...

    2014

  3. [11]

    Graph coloring based physical-cell-ID assignment for LTE networks,

    T. Bandh, G. Carle, and H. Sanneck, “Graph coloring based physical-cell-ID assignment for LTE networks,” in Proc. of the 2009 Int. Conf. Wireless Commun. and Mobile Comput.: Connecting the world wirelessly , ser. IWCMC ’09. Association for Computing Machinery, 2009, p. 116–120...

    2009

  4. [12]

    Randomized graph coloring algorithm for physical cell id assignment in lte-a femtocellular networks,

    A. Pratap, R. Misra, and U. Gupta, “Randomized graph coloring algorithm for physical cell id assignment in lte-a femtocellular networks,” Wireless Pers. Commun. , vol. 91, pp. 1213–1235, 2016, doi:10.1007/s11277-016- 3522-3

    2016 doi

  5. [13]

    PCI planning method based on genetic algorithm in LTE network,

    P. Li and J. Wang, “PCI planning method based on genetic algorithm in LTE network,” Telecom- mun. Sci. , vol. 32, no. 3, p. 2016082, 2016, doi:10.11959/j.issn.1000-0801.2016082

    2016

  6. [14]

    A novel PCI optimization method in LTE system based on intelligent genetic algorithm,

    A. Shen, B. Guo, Y . Gao, T. Xie, X. Hu, Y . Zhang, J. Shen, Y . Fang, G. Wang, and Y . Liu, “A novel PCI optimization method in LTE system based on intelligent genetic algorithm,” in Int. Conf. Signal and Inf. Process., Netw. And Comput. Springer, 2017, pp. 350–355, doi:10.10...

    2017 doi

  7. [15]

    The physical cell identity assignment problem: A practical optimization approach,

    C. E. Andrade, L. S. Pessoa, and S. Stawiarski, “The physical cell identity assignment problem: A practical optimization approach,” IEEE Trans. Evol. Comput. , vol. 28, no. 2, pp. 282–292, 2024, doi:10.1109/TEVC.2022.3185927

    2024

  8. [16]

    A two-level graph partitioning problem arising in mobile wireless communications,

    J. Fairbrother, A. N. Letchford, and K. Briggs, “A two-level graph partitioning problem arising in mobile wireless communications,” Comput. Optim. and Appl. , vol. 69, pp. 653–676, 2018, doi:10.1007/s10589-017- 9967-9

    2018 doi

  9. [17]

    5G PCI and PRACH planning,

    Huawei Technologies Co., Ltd., “5G PCI and PRACH planning,” https://www.scribd.com/document/ 473706035/5G-PCI-and-PRACH-Planning-pdf

  10. [18]

    PCI optimization method and apparatus,

    Q. Zhang, S. Chen, j. Sui, Z. Yao, S. Zhang, Y . Liu, and X. Wang, “PCI optimization method and apparatus,” China Patent CN105 744 548A, 2014

    2014

  11. [19]

    Relation matrix establishment system and method and PCI intelligent optimization system and method,

    M. Ruan, “Relation matrix establishment system and method and PCI intelligent optimization system and method,” China Patent CN103 906 078A, 2014

    2014

  12. [20]

    C. Ding, D. Pei, and A. Salomaa, Chinese remainder theorem: applications in computing, coding, cryptogra- phy. USA: World Scientific Publishing Co., Inc., 1996, doi:10.1142/3254

    1996 doi

  13. [21]

    The max-cut problem and quadratic 0-1 optimization; polyhedral aspects, re- laxations and bounds,

    E. Boros and P. L. Hammer, “The max-cut problem and quadratic 0-1 optimization; polyhedral aspects, re- laxations and bounds,” Ann. Operations Res. , vol. 33, no. 3, pp. 151–180, 1991, doi:10.1007/BF02115753

    1991 doi

  14. [22]

    On the hardness of approximating max k-cut and its dual,

    V . Kann, S. Khanna, J. Lagergren, and A. Panconesi, “On the hardness of approximating max k-cut and its dual,” in Int. Seminar Theor. Comput. Sci. , 1996, pp. 61–67

    1996

  15. [23]

    A multiple search operator heuris- tic for the max-k-cut problem,

    F. Ma and J.-K. Hao, “A multiple search operator heuris- tic for the max-k-cut problem,” Ann. Operations Res. , vol. 248, pp. 365–403, 2017, doi:10.1007/s10479-016- 2234-0

    2017 doi

  16. [24]

    Welsh powell algorithm to color graphs,

    G. G. Subies, “Welsh powell algorithm to color graphs,” https://gist.github.com/GuillemGSubies/ 59b968ec0a68c17384c9f15c15cb38e6

  17. [25]

    Relaxation-free min-k-partition for PCI assignment in 5G networks,

    Y . Qiu, C. Huang, Y . Xue, Z. Jiang, Q. Shi, D. Zhang, and Z.-Q. Luo, “Relaxation-free min-k-partition for PCI assignment in 5G networks,” arXiv preprint arXiv:2506.10362 , 2025, doi:10.48550/arXiv.2506.10362

    2025 doi arXiv

  18. [26]

    Laplacian eigenvalues and the maximum cut problem,

    C. Delorme and S. Poljak, “Laplacian eigenvalues and the maximum cut problem,” Math. Program. , vol. 62, pp. 557–574, 1993, doi:10.1007/BF01585184. JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2015 16

    1993 doi

  19. [27]

    On the shannon capacity of a graph,

    L. Lovász, “On the shannon capacity of a graph,” IEEE Trans. Inf. Theory , vol. 25, no. 1, pp. 1–7, 1979, doi:10.1109/TIT.1979.1055985

    1979

  20. [28]

    Cones of matrices and set- functions and 0–1 optimization,

    L. Lovász and A. Schrijver, “Cones of matrices and set- functions and 0–1 optimization,” SIAM J. Optim., vol. 1, no. 2, pp. 166–190, 1991, doi:10.1137/0801013

    1991 doi

  21. [29]

    Nonpolyhedral relaxations of graph-bisection problems,

    S. Poljak and F. Rendl, “Nonpolyhedral relaxations of graph-bisection problems,” SIAM J. Optim., vol. 5, no. 3, pp. 467–487, 1995, doi:10.1137/0805024

    1995 doi

  22. [30]

    Quadratic optimization problems,

    N. Z. Shor, “Quadratic optimization problems,” Sov. J. Comput. and Syst. Sci. , vol. 25, pp. 1–11, 1987

    1987

  23. [31]

    Rank-two re- laxation heuristics for max-cut and other binary quadratic programs,

    S. Burer, R. D. Monteiro, and Y . Zhang, “Rank-two re- laxation heuristics for max-cut and other binary quadratic programs,” SIAM J. Optim. , vol. 12, no. 2, pp. 503–521, 2002, doi:10.1137/S1052623400382467

    2002 doi

  24. [32]

    Clus- tering by orthogonal nmf model and non-convex penalty optimization,

    S. Wang, T.-H. Chang, Y . Cui, and J.-S. Pang, “Clus- tering by orthogonal nmf model and non-convex penalty optimization,” IEEE Trans. Signal Process. , vol. 69, pp. 5273–5288, 2021, doi:10.1109/TSP.2021.3102106

    2021

  25. [33]

    Nocedal and S

    J. Nocedal and S. J. Wright, Numerical Optimization . Springer, 1999, doi:10.1007/978-0-387-40065-5

    1999 doi

  26. [34]

    Sparse signal estimation by maximally sparse convex optimization,

    I. W. Selesnick and I. Bayram, “Sparse signal estimation by maximally sparse convex optimization,” IEEE Trans. Signal Process. , vol. 62, no. 5, pp. 1078–1092, 2014, doi:10.1109/TSP.2014.2298839

    2014

  27. [35]

    Penalty dual decomposition method for nonsmooth nonconvex optimization—part I: Algorithms and convergence analysis,

    Q. Shi and M. Hong, “Penalty dual decomposition method for nonsmooth nonconvex optimization—part I: Algorithms and convergence analysis,” IEEE Trans. Signal Process., vol. 68, pp. 4108–4122, 2020

    2020

  28. [36]

    Penalty dual decomposition method for nonsmooth noncon- vex optimization—part II: Applications,

    Q. Shi, M. Hong, X. Fu, and T.-H. Chang, “Penalty dual decomposition method for nonsmooth noncon- vex optimization—part II: Applications,” IEEE Trans. Signal Process. , vol. 68, pp. 4242–4257, 2020, doi:10.1109/TSP.2020.3001397

    2020

  29. [37]

    Beck, First-order methods in optimization

    A. Beck, First-order methods in optimization . SIAM, 2017, doi:10.1137/1.9781611974997

    2017 doi

  30. [38]

    On the convergence rate of stochastic mirror descent for nonsmooth nonconvex optimization,

    S. Zhang and N. He, “On the convergence rate of stochastic mirror descent for nonsmooth nonconvex optimization,” arXiv preprint arXiv:1806.04781 , 2018, doi:10.48550/arXiv.1806.04781

    2018 doi arXiv

  31. [39]

    Stochastic mirror descent in variationally coherent optimization problems,

    Z. Zhou, P. Mertikopoulos, N. Bambos, S. Boyd, and P. W. Glynn, “Stochastic mirror descent in variationally coherent optimization problems,” in Advances in Neural Inf. Process. Syst. , vol. 30. Curran Associates, Inc., 2017

    2017

  32. [40]

    Mirror descent in non-convex stochastic programming,

    Z. Zhou, P. Mertikopoulos, N. Bambos, S. Boyd, and P. Glynn, “Mirror descent in non-convex stochastic programming,” arXiv preprint arXiv:1706.05681 , 2017, doi:10.48550/arXiv.1706.05681

    2017 doi arXiv

  33. [41]

    A biased random-key genetic algorithm for wireless backhaul network design,

    C. E. Andrade, M. G. Resende, W. Zhang, R. K. Sinha, K. C. Reichmann, R. D. Doverspike, and F. K. Miyazawa, “A biased random-key genetic algorithm for wireless backhaul network design,” Appl. Soft Comput. , vol. 33, pp. 150–169, 2015, doi:10.1016/j.asoc.2015.04.016

    2015 doi

  34. [42]

    Communi- cation over a wireless network with random connections,

    R. Gowaikar, B. Hochwald, and B. Hassibi, “Communi- cation over a wireless network with random connections,” IEEE Trans. Inf. Theory , vol. 52, no. 7, pp. 2857–2871, 2006, doi:10.1109/TIT.2006.876254

    2006

  35. [43]

    Stochastic geometry and random graphs for the analysis and design of wireless networks,

    M. Haenggi, J. G. Andrews, F. Baccelli, O. Dousse, and M. Franceschetti, “Stochastic geometry and random graphs for the analysis and design of wireless networks,” IEEE J. Sel. Areas Commun. , vol. 27, no. 7, pp. 1029– 1046, 2009, doi:10.1109/JSAC.2009.090902. JOURNAL OF LATEX ...

    2009 arXiv

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.