Pith. sign in

REVIEW 4 major objections 6 minor 2 cited by

When networks are optimized under budgets and dynamical goals, classic features like sparsity, bipartiteness, and community splits appear on their own instead of being imposed.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · grok-4.5

2026-07-15 12:08 UTC pith:DYBV2F7S

load-bearing objection Solid methods paper: autodiff-through-dynamics network design with real constraint encoding and a few clean recoveries; the unification rhetoric and continuous-to-discrete transfer are the soft spots, not the core engineering contribution. the 4 major comments →

arxiv 2603.09197 v2 pith:DYBV2F7S submitted 2026-03-10 physics.soc-ph nlin.AO

GradNet: A Gradient-Based Framework for Optimal Network Science

Guram Mikaberidze , Beso Mikaberidze , Dane Taylor This is my paper
classification physics.soc-ph nlin.AO
keywords network optimizationGradNetsynchronizationKuramoto modelalgebraic connectivityopinion dynamicsquantum networksvariational network science
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

Network science has long treated topology as given and asked how it shapes dynamics. This paper inverts that: it treats the network itself as the decision variable and asks how functional goals and resource limits produce architecture. GradNet turns adjacency matrices into smooth, constrained objects so first-order optimizers can design networks for arbitrary differentiable objectives—synchronization, social tension, quantum communication capacity—while respecting budgets, geometry, and sign rules. Across case studies the same pattern appears: sparsity, bipartite frequency-disassortative wiring, factional splits, and minimum spanning trees emerge without being coded in. The framework is both an engineering tool that scales past 100,000 nodes and a scientific probe that turns structure–function questions into constrained optimization problems.

Core claim

Constrained gradient optimization of network topology under realistic resource, geometric, and sign constraints spontaneously produces canonical structural features—sparsity, bipartiteness, community partition, and tree-like backbones—rather than requiring them to be imposed. Optimizing Kuramoto synchrony under a fixed coupling budget yields sparse, bipartite, frequency-disassortative networks that eliminate classical synchronization thresholds; minimizing social tension on Zachary’s karate club recovers the observed factional split; maximizing spatial quantum capacity under distance-dependent costs recovers the minimum spanning tree.

What carries the argument

GradNet: a smooth encoding that maps unconstrained parameters through symmetry, edge masks, budget norms, and sign constraints into a valid adjacency matrix, so automatic differentiation can push gradients of static or dynamical losses straight through the topology (including through numerical ODE integration).

Load-bearing premise

That continuous smooth parameterizations of adjacency matrices plus first-order optimizers reliably reach architectures that are meaningful optima of the intended discrete design problem, not artifacts of the relaxation or of local minima.

What would settle it

For the Kuramoto case: from many random initializations and optimizers under the same coupling budget, check whether recovered networks systematically lack bipartiteness or monophily, or retain a finite synchronization threshold; if they do, the spontaneous-emergence claim fails.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • Network design for synchrony, diffusion, and quantum capacity can be run at scales exceeding 10^5 nodes with GPU gradient methods.
  • Sparsity and bipartiteness can be predicted as outcomes of resource-limited objectives rather than assumed as primitives.
  • Optimal Kuramoto networks remove classical synchronization thresholds and admit reduced continuum descriptions.
  • Opinion-dynamics rewiring under social tension can act as a generative model of community structure.
  • Optimization plus variational analysis can convert otherwise intractable network dynamics into partially solvable problems.

Where Pith is reading between the lines

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

  • The same continuous encoding can unify forward design and inverse inference (recovering topology from trajectories) under one codebase and theory.
  • If real biological or infrastructure networks sit near these optima, measured modularity and degree patterns could be reverse-engineered as signatures of particular budgets and objectives.
  • Bottleneck and synchrony losses both pressure networks toward sparse spanning or bipartite skeletons, suggesting a broader principle that many ‘natural’ graph families are critical points of simple resource-limited objectives.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 6 minor

Summary. The manuscript introduces GradNet, a first-order optimization framework that parameterizes admissible network topologies (directed/undirected, sign and mask constraints, ℓp resource budgets) as smooth maps from unconstrained parameters, enabling automatic differentiation through static spectral objectives and through discrete- or continuous-time dynamics. Across case studies—logical gates, algebraic connectivity on lattices, Kuramoto synchrony under fixed coupling budgets, opinion-driven pruning of Zachary’s karate club, reconstruction from phase trajectories, and spatial quantum capacity under distance costs—the authors report that sparsity, bipartiteness, community splits, and minimum-spanning-tree backbones arise without being imposed, and they present a closed-form optimum for lattice λ₂ (Eq. 8) matching numerics. The paper positions constrained optimization as both an engineering design tool (claimed scalable beyond 10^5 nodes) and a scientific probe of structure–function relations.

Significance. If the continuous optima reliably correspond to meaningful discrete design solutions and the emergence results generalize, this is a useful methodological contribution: a single, software-backed pipeline that optimizes arbitrary differentiable dynamical objectives under realistic constraints, with demonstrated GPU/sparse scaling and a package release. The lattice algebraic-connectivity analysis (variational conditions → closed form Eq. 8 vs Fig. 3) is a concrete example of theory–computation interplay. The Kuramoto and quantum case studies, if robust, would strengthen the claim that optimization can surface non-obvious architectures (bipartite monophilic networks; MST under bottleneck capacity). Related spectral and rewiring work is acknowledged; the main advance is breadth of objectives (including full ODE integration) and the unified encoding rather than a wholly new variational idea.

major comments (4)
  1. §§II–III and Abstract/§V: The central inference—that GradNet solutions show canonical features “emerge spontaneously” and that the method is a reliable engineering tool—requires that continuous local (or noise-smoothed) optima of the soft-parameterized map (Steps 1–6: softabs, masks, budget rescaling) transfer to the intended discrete design problem. Only the lattice λ₂ case supplies an independent closed form (Eq. 8). For Kuramoto (§IV-C), karate (§IV-D), reconstruction (§IV-E), and quantum capacity (§IV-F), the manuscript reports single gradient trajectories without multi-start statistics, discrete rewiring/SA/convex baselines, or post-hoc hard-threshold fidelity. Without that evidence, “emergent” sparsity/bipartiteness/MST and the unification claim remain under-supported.
  2. Abstract and §I claim scalability “exceeding 10^5 nodes.” Fig. 3(d) and the accompanying text show sparse vs dense timing for algebraic connectivity and a 100×100 lattice (~10^4 nodes, 2.7 h on L40S). No experiment, memory profile, or wall-clock result at N>10^5 is reported for a dynamics-dependent loss (which requires repeated ODE integration). Either provide such a result for at least one dynamical objective or qualify the claim to the spectral/sparse regime actually demonstrated.
  3. §IV-C (Kuramoto): The strong dynamical claim that optimized networks “eliminate classical synchronization thresholds” and exhibit universal critical scaling is largely deferred to prior work [29]. For a self-contained journal article advertising this as a flagship GradNet result, the manuscript should include (i) order-parameter vs budget curves for optimized vs standard ensembles, (ii) a clear statement of what is new here versus [29], and (iii) at least a brief multi-seed or discrete-search check that the sparse bipartite monophilic architecture is not a local-minimum artifact of the continuous encoding.
  4. §IV-D (karate club): The setup fixes two opposing opinions and allows only nonpositive edge modifications (friendship deletion) while minimizing long-time social tension (Eqs. 16–17). Under those constraints a bipartition is strongly favored; recovering the known split (one ambiguous node) is interesting but does not by itself establish that community structure “emerges” from unconstrained optimization. Clarify the design choices as a pruning/inference experiment rather than a general generative principle, or add a control (e.g., allowing edge addition or random fixed opinions) showing the split is not forced by the sign/mask constraints alone.
minor comments (6)
  1. Table II is useful but dense; several cells use “N/A” inconsistently with the prose (e.g., masks for Kuramoto). A short caption note defining each column would help.
  2. Fig. 4 (Kuramoto snapshots) is described as becoming sparse/bipartite/monophilic/elongated, but quantitative diagnostics (edge density, bipartiteness index, frequency assortativity, diameter) over optimization time are not plotted; adding them would make the emergence claim easier to assess.
  3. §III soft-absolute approximation and renormalization ||θ||_F=√m are important hyperparameters; default values and sensitivity should be stated once for reproducibility.
  4. Related-work paragraph on spectral objectives and zeroth-order rewiring is good; a short explicit comparison table (objective class, continuous vs discrete, max N, dynamics-through-AD?) would sharpen novelty relative to [24]–[37] and GNN rewiring [45],[46].
  5. Minor notation: adjacency is written both A_ij and A; Laplacian action in Eq. (3) uses double indices that could be defined once more carefully for readers outside spectral graph theory.
  6. Package URL [48] is welcome; stating license, version, and which figure scripts are released would strengthen the engineering claim.

Circularity Check

2 steps flagged

No definitional circularity in the core GradNet claims; mild setup bias only in the karate and quantum demonstrations where the chosen loss+constraints already favor bipartition/MST.

specific steps
  1. other [§IV (Zachary’s Karate Club paragraph) and Abstract]
    "Minimizing social tension in opinion dynamics reproduces the empirically observed factional split in Zachary’s karate club network. ... The opinions of the president and the instructor are fixed x_p(t)=+1, x_i(t)=−1. ... The final edges are restricted to be non-negative, and the allowed modifications are non-positive ... The loss function is defined as the long-time average of the social tension."

    With two fixed opposing opinion sources and only edge-deletion allowed, the tension loss ½∑A_ij(x_i−x_j)² is minimized precisely by severing cross-basin edges. The resulting bipartition is therefore a near-direct attractor of the chosen dynamics+constraints rather than an independent structural discovery; the match to the historical factions is largely a consequence of planting the two poles and the original community structure.

  2. other [§IV-F (quantum internet) and Abstract]
    "Maximizing entanglement distribution in spatial quantum networks under distance-dependent costs recovers minimum spanning tree architectures. ... Optimizing the quantum communication capacity in Eq. (22) subject to ... recovers the minimum spanning tree as the optimal network architecture (Fig. 7). This outcome is intuitive: ... The resulting optimum is therefore the minimum spanning tree."

    The nested min–max form of the mean capacity (bottleneck transmissivity along the best path) under a linear distance budget already implies, by the authors’ own analysis, that redundant edges cannot help and that the optimum must be the MST. The numerical recovery is therefore confirmatory of the objective’s geometry rather than a spontaneous emergence discovered by GradNet.

full rationale

The paper's central results are obtained by running first-order optimization of externally defined, differentiable objectives (Kuramoto order parameter, algebraic connectivity λ₂, social tension, prediction error, mean quantum capacity) subject to explicit resource/sign/mask constraints. The optimized architectures are not algebraically identical to the inputs by construction, nor are they forced by a uniqueness theorem imported from the authors. The lattice λ₂ case supplies an independent closed-form derivation (Eq. 8) that matches numerics. The Kuramoto structures and threshold elimination are numerical outcomes (reproduced from the authors' prior work [29]) that are then analyzed, not defined into the loss. Reconstruction recovers a known ground-truth adjacency from trajectories without tautology. The only mild circularity risks are (i) the karate-club experiment, whose fixed opposing leaders plus non-positive edge updates make a bipartition a natural attractor of the tension loss, and (ii) the quantum case, where the nested min–max capacity objective under a distance budget is shown by the authors themselves to select the MST, so the optimizer 'recovering' the MST is largely confirmatory. These are setup choices, not self-definitional reductions of the claimed variational principle. Self-citations ([29], package [48]) are normal and not load-bearing for uniqueness. Score 2 reflects only those two minor, non-central attractors; the framework and remaining case studies stand as non-circular.

Axiom & Free-Parameter Ledger

5 free parameters · 5 axioms · 1 invented entities

The central claim rests on standard autodiff/optimization math, standard dynamical models (Kuramoto, diffusion, quantum transmissivity maps), and modeling choices that encode budgets and masks. Free parameters are optimizer and smoothing knobs, not physics constants fitted to claim the emergence results. No new physical entities are postulated; GradNet is a computational method. Load-bearing domain assumptions are that continuous relaxations plus local gradient steps adequately represent discrete network design under the stated constraints.

free parameters (5)
  • learning rate α and optimizer hyperparameters (e.g., Adam)
    Chosen for training stability; affect which local optimum is reached in non-convex losses (§III).
  • soft-absolute smoothing ε
    Smooth |·| approximation for sign constraints; can change the loss landscape near zero weights (§III Step 6).
  • noise radius r and batch size n (stochastic smoothing)
    Hand-tuned mollification for non-smooth quantum capacity objective; required for MST recovery narrative (§IV-F).
  • budget B and cost matrix C_ij
    Problem inputs that define the admissible set; different B/C change emergent sparsity and topology (Eqs. 2, 14, 19).
  • parameter renormalization target ||θ||_F = √m
    Default rescaling after each step to stabilize gradients under exact budget; affects optimization dynamics though not the forward map (§III).
axioms (5)
  • standard math Automatic differentiation correctly propagates gradients through numerical ODE integration and spectral/Rayleigh computations used as losses.
    Assumed throughout §§II–IV; standard in differentiable programming but sensitive to integrator and eigenvalue multiplicity.
  • domain assumption Admissible networks are faithfully represented by the six-step smooth map from unconstrained θ to A (symmetry, signs, mask, budget, weight signs).
    §III pipeline; conflicts between modification-sign and weight-sign constraints are acknowledged as imperfect.
  • domain assumption Kuramoto, Laplacian diffusion, and bottleneck transmissivity models adequately proxy the real systems being designed or explained.
    Case studies §§IV-C–F; standard models but not validated against field measurements here.
  • standard math For algebraic connectivity on the grid, equal marginal gains of λ2 w.r.t. each edge weight characterize the constrained optimum (Eq. 5).
    Variational first-order optimality under budget; used to derive closed form Eq. 8.
  • ad hoc to paper Sparsity and other structural features need not be imposed if they maximize the chosen objective under budget.
    Interpretive stance of Abstract/§I/§V; true only relative to the selected losses and constraints.
invented entities (1)
  • GradNet smooth admissible-set encoding (dense/sparse) independent evidence
    purpose: Map free parameters to constraint-satisfying adjacency matrices so first-order optimizers can design networks.
    Methodological construct, not a physical entity; independent evidence is the open-source package and case-study recoveries, not an external measurement.

pith-pipeline@v1.1.0-grok45 · 22251 in / 3691 out tokens · 39915 ms · 2026-07-15T12:08:45.615376+00:00 · methodology

0 comments
read the original abstract

Network science has traditionally examined how structure determines dynamics. Here we invert this paradigm: we ask how functional dynamics and resource constraints shape network architecture. We introduce GradNet, an AI-enabled optimization framework that treats network topology as a continuously differentiable object. This allows designing networks that optimize arbitrary dynamical objectives, from synchronization to communication capacity, under realistic constraints. Applying this framework across diverse systems reveals that canonical network features emerge spontaneously from constrained optimization rather than requiring explicit imposition. Optimizing Kuramoto oscillator synchronization under fixed coupling budgets produces sparse, bipartite, frequency-disassortative architectures that eliminate classical synchronization thresholds. Minimizing social tension in opinion dynamics reproduces the empirically observed factional split in Zachary's karate club network. Maximizing entanglement distribution in spatial quantum networks under distance-dependent costs recovers minimum spanning tree architectures. These results demonstrate that optimization acts as both an engineering tool for network design, scalable to networks exceeding $10^5$ nodes, and a scientific probe revealing fundamental structure-function relationships. By recasting network architecture as the solution to constrained optimization problems, this variational perspective offers a unified framework connecting network analysis, design, and inference across physical, biological, and technological systems.

discussion (0)

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

Forward citations

Cited by 2 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Mapping open quantum dynamics onto graphs

    quant-ph 2026-07 accept novelty 7.0

    Markovian quantum master equations are exactly equivalent to averaged wave features of operator-valued signals on two uniquely defined magnetic graphs with vertex potentials.

  2. A Multi-Agent Framework for Zero-Dimensional Reduced-Order Model Planning

    cs.LG 2026-07 conditional novelty 6.0

    A multi-agent LLM framework with ontology RAG and MILP-guided search automates forward and inverse 0D reduced-order network design across aero-engine air systems, power grids, and water networks.

Reference graph

Works this paper leans on

69 extracted references · cited by 2 Pith papers

  1. [1]

    Statistical mechanics of complex networks,

    R. Albert and A.-L. Barab´asi, “Statistical mechanics of complex networks,” ������� �� ������ �������, vol. 74, no. 1, p. 47, 2002

    2002

  2. [2]

    Newman,��������

    M. Newman,��������. Oxford university press, 2018

    2018

  3. [3]

    The structure and function of complex networks,

    M. E. Newman, “The structure and function of complex networks,”���� ������, vol. 45, no. 2, pp. 167–256, 2003

    2003

  4. [4]

    Complex networks: Structure and dynamics,

    S. Boccaletti, V . Latora, Y . Moreno, M. Chavez, and D.-U. Hwang, “Complex networks: Structure and dynamics,”������� �������, vol. 424, no. 4-5, pp. 175–308, 2006

    2006

  5. [5]

    Synchronization in complex networks,

    A. Arenas, A. D ´ıaz-Guilera, J. Kurths, Y . Moreno, and C. Zhou, “Synchronization in complex networks,”������� �������, vol. 469, no. 3, pp. 93–153, 2008

    2008

  6. [6]

    Barrat, M

    A. Barrat, M. Barthelemy, and A. Vespignani,��������� ��������� �� ������� ��������. Cambridge university press, 2008

    2008

  7. [7]

    Epidemic processes in complex networks,

    R. Pastor-Satorras, C. Castellano, P. Van Mieghem, and A. Vespignani, “Epidemic processes in complex networks,”������� �� ������ �������, vol. 87, no. 3, pp. 925–979, 2015

    2015

  8. [8]

    On random graphs i,

    P. Erd˝os and A. R ´enyi, “On random graphs i,”������������� ���������� ���, vol. 6, pp. 290–297, 1959

    1959

  9. [9]

    Collective dynamics of ‘small- world’networks,

    D. J. Watts and S. H. Strogatz, “Collective dynamics of ‘small- world’networks,”������, vol. 393, no. 6684, pp. 440–442, 1998

    1998

  10. [10]

    Highly optimized tolerance: Robustness and design in complex systems,

    J. M. Carlson and J. Doyle, “Highly optimized tolerance: Robustness and design in complex systems,”�������� ������ �������, vol. 84, no. 11, p. 2529, 2000

    2000

  11. [11]

    Size and form in efficient transportation networks,

    J. R. Banavar, A. Maritan, and A. Rinaldo, “Size and form in efficient transportation networks,”������, vol. 399, no. 6732, pp. 130–132, 1999

    1999

  12. [12]

    Nonoptimal component placement, but short processing paths, due to long-distance projections in neural systems,

    M. Kaiser and C. C. Hilgetag, “Nonoptimal component placement, but short processing paths, due to long-distance projections in neural systems,” ���� ������������� �������, vol. 2, no. 7, p. e95, 2006

    2006

  13. [13]

    Fluctuations and redundancy in optimal transport networks,

    F. Corson, “Fluctuations and redundancy in optimal transport networks,” �������� ������ �������, vol. 104, no. 4, p. 048703, 2010

    2010

  14. [14]

    Optimization in complex networks,

    R. F. I. Cancho and R. V . Sol ´e, “Optimization in complex networks,” in ����������� ��������� �� ������� ��������. Springer, 2003, pp. 114–126

    2003

  15. [15]

    The best is yet to come,

    M. Buchanan, “The best is yet to come,”������, vol. 447, no. 7140, pp. 39–39, 2007

    2007

  16. [16]

    Phase transitions in pareto optimal complex networks,

    L. F. Seoane and R. Sol ´e, “Phase transitions in pareto optimal complex networks,”�������� ������ �, vol. 92, no. 3, p. 032807, 2015

    2015

  17. [17]

    Mitigation of malicious attacks on networks,

    C. M. Schneider, A. A. Moreira, J. S. Andrade Jr, S. Havlin, and H. J. Herrmann, “Mitigation of malicious attacks on networks,”����������� �� ��� �������� ������� �� ��������, vol. 108, no. 10, pp. 3838–3841, 2011

    2011

  18. [18]

    Onion-like network topology enhances robustness against malicious attacks,

    H. J. Herrmann, C. M. Schneider, A. A. Moreira, J. S. Andrade Jr, and S. Havlin, “Onion-like network topology enhances robustness against malicious attacks,”������� �� ����������� ���������� ������ ��� ����������, vol. 2011, no. 01, p. P01027, 2011

    2011

  19. [19]

    Enhancing network robustness against malicious attacks,

    A. Zeng and W. Liu, “Enhancing network robustness against malicious attacks,”�������� ������ �������������� ���������� ��� ���� ������ �������, vol. 85, no. 6, p. 066130, 2012

    2012

  20. [20]

    Smart rewiring for network robustness,

    V . H. Louzada, F. Daolio, H. J. Herrmann, and M. Tomassini, “Smart rewiring for network robustness,”������� �� ������� ��������, vol. 1, no. 2, pp. 150–159, 2013

    2013

  21. [21]

    Improving the robustness of complex networks with preserving community structure,

    Y . Yang, Z. Li, Y . Chen, X. Zhang, and S. Wang, “Improving the robustness of complex networks with preserving community structure,” ���� ���, vol. 10, no. 2, p. e0116551, 2015

    2015

  22. [22]

    Optimizing network robustness by edge rewiring: a general framework,

    H. Chan and L. Akoglu, “Optimizing network robustness by edge rewiring: a general framework,”���� ������ ��� ��������� ���������, vol. 30, no. 5, pp. 1395–1425, 2016

    2016

  23. [23]

    Optimizing network robustness via krylov subspaces,

    S. Massei and F. Tudisco, “Optimizing network robustness via krylov subspaces,”������ ������������ ��������� ��� ��������� ��������, vol. 58, no. 1, pp. 131–155, 2024

    2024

  24. [24]

    Network synchronization, diffusion, and the paradox of heterogeneity,

    A. E. Motter, C. Zhou, and J. Kurths, “Network synchronization, diffusion, and the paradox of heterogeneity,”�������� ������ �������������� ���������� ��� ���� ������ �������, vol. 71, no. 1, p. 016116, 2005

    2005

  25. [25]

    Entangled networks, synchronization, and optimal network topology,

    L. Donetti, P. I. Hurtado, and M. A. Munoz, “Entangled networks, synchronization, and optimal network topology,”�������� ������ �������, vol. 95, no. 18, p. 188701, 2005

    2005

  26. [26]

    Efficient rewirings for enhancing synchronizability of dynamical networks,

    A. A. Rad, M. Jalili, and M. Hasler, “Efficient rewirings for enhancing synchronizability of dynamical networks,”������ �� ����������������� ������� �� ��������� �������, vol. 18, no. 3, 2008

    2008

  27. [27]

    Evolving enhanced topologies for the synchronization of dynamical complex networks,

    T. E. Gorochowski, M. di Bernardo, and C. S. Grierson, “Evolving enhanced topologies for the synchronization of dynamical complex networks,”�������� ������ �������������� ���������� ��� ���� ������ �������, vol. 81, no. 5, p. 056212, 2010

    2010

  28. [28]

    Synchronization of heterogeneous oscillators under network modifications: Perturbation and optimization of the synchrony alignment function,

    D. Taylor, P. S. Skardal, and J. Sun, “Synchronization of heterogeneous oscillators under network modifications: Perturbation and optimization of the synchrony alignment function,”���� ������� �� ������� �����������, vol. 76, no. 5, pp. 1984–2008, 2016

    1984

  29. [29]

    Emergent topology of optimal networks for synchrony,

    G. Mikaberidze and D. Taylor, “Emergent topology of optimal networks for synchrony,”����� �������� ����������������, 2025

    2025

  30. [30]

    Fastest mixing markov chain on a graph,

    S. Boyd, P. Diaconis, and L. Xiao, “Fastest mixing markov chain on a graph,”���� ������, vol. 46, no. 4, pp. 667–689, 2004

    2004

  31. [31]

    Approximation algorithms for reducing the spectral radius to control epidemic spread,

    S. Saha, A. Adiga, B. A. Prakash, and A. K. S. Vullikanti, “Approximation algorithms for reducing the spectral radius to control epidemic spread,” in����������� �� ��� ���� ���� ������������� ���������� �� ���� ������. SIAM, 2015, pp. 568–576

    2015

  32. [32]

    Met: A fast algorithm for minimizing propagation in large graphs with small eigen-gaps,

    L. T. Le, T. Eliassi-Rad, and H. Tong, “Met: A fast algorithm for minimizing propagation in large graphs with small eigen-gaps,” in ����������� �� ��� ���� ���� ������������� ���������� �� ���� ������. SIAM, 2015, pp. 694–702

    2015

  33. [33]

    Distributed topology manipulation to control epidemic spreading over networks,

    D. Xue and S. Hirche, “Distributed topology manipulation to control epidemic spreading over networks,”���� ������������ �� ������ ����������, vol. 67, no. 5, pp. 1163–1174, 2018

    2018

  34. [34]

    Optimizing controllability of complex networks by minimum structural perturbations,

    W.-X. Wang, X. Ni, Y .-C. Lai, and C. Grebogi, “Optimizing controllability of complex networks by minimum structural perturbations,”�������� ������ �������������� ���������� ��� ���� ������ �������, vol. 85, no. 2, p. 026115, 2012

    2012

  35. [35]

    Network design for controllability metrics,

    C. O. Becker, S. Pequito, G. J. Pappas, and V . M. Preciado, “Network design for controllability metrics,”���� ������������ �� ������� �� ������� �������, vol. 7, no. 3, pp. 1404–1415, 2020

    2020

  36. [36]

    Growing well-connected graphs,

    A. Ghosh and S. Boyd, “Growing well-connected graphs,” in����������� �� ��� ���� ���� ���������� �� �������� ��� �������. IEEE, 2006, pp. 6605–6611

    2006

  37. [37]

    Minimizing effective resistance of a graph,

    A. Ghosh, S. Boyd, and A. Saberi, “Minimizing effective resistance of a graph,”���� ������, vol. 50, no. 1, pp. 37–66, 2008

    2008

  38. [38]

    Symmetry breaking in optimal transport networks,

    S. Patwardhan, M. Barthelemy, S ¸. Erkol, S. Fortunato, and F. Radicchi, “Symmetry breaking in optimal transport networks,”������ ���������� �����, vol. 15, no. 1, p. 3758, 2024

    2024

  39. [39]

    Optimal design of spatial distribution networks,

    M. T. Gastner and M. E. Newman, “Optimal design of spatial distribution networks,”�������� ������ �������������� ���������� ��� ���� ������ �������, vol. 74, no. 1, p. 016117, 2006

    2006

  40. [40]

    Master stability functions for synchro- nized coupled systems,

    L. M. Pecora and T. L. Carroll, “Master stability functions for synchro- nized coupled systems,”�������� ������ �������, vol. 80, no. 10, p. 2109, 1998

    1998

  41. [41]

    Epidemic spreading in real networks: An eigenvalue viewpoint,

    Y . Wang, D. Chakrabarti, C. Wang, and C. Faloutsos, “Epidemic spreading in real networks: An eigenvalue viewpoint,” in���� ������������� ��������� �� �������� ����������� �������� ����� ������������IEEE, 2003, pp. 25–34

    2003

  42. [42]

    Optimal network topologies: expanders, cages, ramanujan graphs, entangled networks and all that,

    L. Donetti, F. Neri, and M. A. Munoz, “Optimal network topologies: expanders, cages, ramanujan graphs, entangled networks and all that,” ������� �� ����������� ���������� ������ ��� ����������, vol. 2006, no. 08, pp. P08 007–P08 007, 2006

    2006

  43. [43]

    Spectral measure of structural robustness in complex networks,

    J. Wu, M. Barahona, Y .-J. Tan, and H.-Z. Deng, “Spectral measure of structural robustness in complex networks,”���� ������������ �� �������� ���� ��� ���������������� �� ������� ��� ������, vol. 41, no. 6, pp. 1244–1252, 2011

    2011

  44. [44]

    Optimal synchronization of complex networks,

    P. S. Skardal, D. Taylor, and J. Sun, “Optimal synchronization of complex networks,”�������� ������ �������, vol. 113, no. 14, p. 144101, 2014

    2014

  45. [45]

    Graph neural networks with adaptive structures,

    Z. Zhang, S. Lu, Z. Huang, and Z. Zhao, “Graph neural networks with adaptive structures,”���� ������� �� �������� ������ �� ������ ����������, vol. 19, no. 1, pp. 181–194, 2024

    2024

  46. [46]

    Aro-gnn: Adaptive relation-optimized graph neural networks,

    Y . Lu and Z. Lin, “Aro-gnn: Adaptive relation-optimized graph neural networks,”������� �� ���� ���� ���������� �������� ��� ����������� ��������, vol. 37, no. 7, p. 174, 2025

    2025

  47. [47]

    Automatic differentiation in machine learning: a survey,

    A. G. Baydin, B. A. Pearlmutter, A. A. Radul, and J. M. Siskind, “Automatic differentiation in machine learning: a survey,”������� �� ������� �������� ��������, vol. 18, no. 153, pp. 1–43, 2018

    2018

  48. [48]

    Gradnet: A python package for optimal network science,

    G. Mikaberidze, B. Mikaberidze, and D. Taylor, “Gradnet: A python package for optimal network science,” https://github.com/mikaberidze/ gradnet, 2026

    2026

  49. [49]

    Algebraic connectivity of graphs,

    M. Fiedler, “Algebraic connectivity of graphs,”������������ ���������� ��� �������, vol. 23, no. 2, pp. 298–305, 1973

    1973

  50. [50]

    Old and new results on algebraic connectivity of graphs,

    N. M. M. De Abreu, “Old and new results on algebraic connectivity of graphs,”������ ������� ��� ��� ������������, vol. 423, no. 1, pp. 53–73, 2007

    2007

  51. [51]

    Maximum algebraic connectivity augmentation is np-hard,

    D. Mosk-Aoyama, “Maximum algebraic connectivity augmentation is np-hard,”���������� �������� �������, vol. 36, no. 6, pp. 677–679, 2008

    2008

  52. [52]

    Optimal robust network design: Formulations and algorithms for maximizing algebraic connec- tivity,

    N. Somisetty, H. Nagarajan, and S. Darbha, “Optimal robust network design: Formulations and algorithms for maximizing algebraic connec- tivity,”���� ������������ �� ������� �� ������� �������, vol. 12, no. 1, pp. 918–929, 2024

    2024

  53. [53]

    ¨Uber die abgrenzung der eigenwerte einer matrix,

    S. A. Gershgorin, “ ¨Uber die abgrenzung der eigenwerte einer matrix,” ���� ����� ���� ���� ���� ����, no. 6, pp. 749–754, 1931

    1931

  54. [54]

    International symposium on mathematical problems in theoretical physics,

    Y . Kuramoto, “International symposium on mathematical problems in theoretical physics,”������� ����� �� �������, vol. 30, p. 420, 1975

    1975

  55. [55]

    From Kuramoto to Crawford: exploring the onset of synchronization in populations of coupled oscillators,

    S. H. Strogatz, “From Kuramoto to Crawford: exploring the onset of synchronization in populations of coupled oscillators,”������� �� ��������� ���������, vol. 143, no. 1-4, pp. 1–20, 2000

    2000

  56. [56]

    Synchronization on small-world networks,

    H. Hong, M.-Y . Choi, and B. J. Kim, “Synchronization on small-world networks,”�������� ������ �, vol. 65, no. 2, p. 026139, 2002

    2002

  57. [57]

    Heterogeneity in oscillator networks: Are smaller worlds easier to synchronize?

    T. Nishikawa, A. E. Motter, Y .-C. Lai, and F. C. Hoppensteadt, “Heterogeneity in oscillator networks: Are smaller worlds easier to synchronize?”�������� ������ �������, vol. 91, no. 1, p. 014101, 2003

    2003

  58. [58]

    On the topology of synchrony optimized networks of a kuramoto-model with non-identical oscillators,

    D. Kelly and G. A. Gottwald, “On the topology of synchrony optimized networks of a kuramoto-model with non-identical oscillators,”������ �� ����������������� ������� �� ��������� �������, vol. 21, no. 2, 2011

    2011

  59. [59]

    An information flow model for conflict and fission in small groups,

    W. W. Zachary, “An information flow model for conflict and fission in small groups,”������� �� ��������������� ��������, vol. 33, no. 4, pp. 452–473, 1977

    1977

  60. [60]

    Community detection in graphs,

    S. Fortunato, “Community detection in graphs,”������� �������, vol. 486, no. 3-5, pp. 75–174, 2010

    2010

  61. [61]

    Finding and evaluating community structure in networks,

    M. E. Newman and M. Girvan, “Finding and evaluating community structure in networks,”�������� ������ �, vol. 69, no. 2, p. 026113, 2004

    2004

  62. [62]

    Infer- ring the connectivity of coupled oscillators from time-series statistical similarity analysis,

    G. Tirabassi, R. Sevilla-Escoboza, J. M. Buld ´u, and C. Masoller, “Infer- ring the connectivity of coupled oscillators from time-series statistical similarity analysis,”��������� �������, vol. 5, no. 1, p. 10829, 2015

    2015

  63. [63]

    Reconstruction of a random phase dynamics network from observations,

    A. Pikovsky, “Reconstruction of a random phase dynamics network from observations,”������� ������� �, vol. 382, no. 4, pp. 147–152, 2018

    2018

  64. [64]

    Inferring networks from time series: A neural approach,

    T. Gaskin, G. A. Pavliotis, and M. Girolami, “Inferring networks from time series: A neural approach,”���� �����, vol. 3, no. 4, p. pgae063, 2024

    2024

  65. [65]

    The quantum internet,

    H. J. Kimble, “The quantum internet,”������, vol. 453, no. 7198, pp. 1023–1030, 2008

    2008

  66. [66]

    Quantum internet: A vision for the road ahead,

    S. Wehner, D. Elkouss, and R. Hanson, “Quantum internet: A vision for the road ahead,”�������, vol. 362, no. 6412, p. eaam9288, 2018

    2018

  67. [67]

    The path to scalable distributed quantum computing,

    R. Van Meter and S. J. Devitt, “The path to scalable distributed quantum computing,”��������, vol. 49, no. 9, pp. 31–42, 2016

    2016

  68. [68]

    Entanglement availability differentiation service for the quantum internet,

    L. Gyongyosi and S. Imre, “Entanglement availability differentiation service for the quantum internet,”��������� �������, vol. 8, no. 1, p. 10620, 2018

    2018

  69. [69]

    End-to-end capacities of a quantum communication network,

    S. Pirandola, “End-to-end capacities of a quantum communication network,”�������������� �������, vol. 2, no. 1, p. 51, 2019

    2019