Pith. sign in

REVIEW 3 major objections 3 minor 46 references

Timescale Separation Through the Lens of Operator Theory

T0 review · 3 major / 3 minor · reviewed 2026-08-04 · deepseek-v4-flash

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

Pith's one-line read This paper proves that an interconnected fast-slow discrete-time system converges linearly whenever the timescale separation parameter δ is smaller than an explicit threshold built from the contraction rates and Lipschitz constants of its f

desk verdict The theorems are correct and the explicit δ̄ bounds are a real contribution, but the paper mislabels its fast-operator assumption as paracontraction when it actually requires a global uniform contraction rate, which narrows the advertised scope. read the letter →

arxiv 2608.02443 v1 pith:UMNMYKL4 submitted 2026-08-03 math.OC cs.SYeess.SY

Guido Carnevale , Nicola Bastianello , Luca Schenato , Giuseppe Notarstefano , Ruggero Carli This is my paper · ORCID
classification math.OCcs.SYeess.SY MSC 93C7047H10
keywords timescaleseparationsingularperturbationoperatortheoryfixed-pointiterationsdiscrete-timesystemsstochasticoperatorsfeedbackoptimizationlinearconvergence
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 bridges singular perturbation ideas and operator theory for discrete-time systems. It considers a slow update x_{t+1}=x_t+δS(x_t,z_t) coupled to a fast update z_{t+1}=F(x_t,z_t), where for each frozen slow state the fast operator has a fixed-point manifold Z_FIX(x). The authors show that if the fast map contracts uniformly toward that manifold and the reduced slow map (with z pinned to the manifold) contracts, then the full interconnection converges linearly for every δ below an explicit threshold δ̄ built from the operator constants. The same mechanism extends to averaged and metrically subregular reduced maps, and to stochastic counterparts with almost sure convergence. The threshold is practical: it turns a qualitative separation assumption into a number one can compute and use to tune a controller.

What carries the argument

The workhorse is a 2×2 comparison matrix M(δ) that tracks two errors jointly: the slow-state error measured as distance to the reduced fixed-point set, and the fast-state error measured as distance to the fast fixed-point manifold. One-step bounds on each error produce the linear recurrence e_{t+1} ≤ M(δ)e_t, and the formula for δ̄ is precisely the condition that M(δ) have spectral radius below one. For the broader case, a lemma converts averagedness plus metric subregularity into an explicit contraction factor, so the same matrix argument applies without requiring the reduced map to be a strict contraction.

What would settle it

Instantiate the linear comparison system at the core of the proof—x_{t+1}=(1−δc_S)x_t+δL_S e_t and e_{t+1}=δL_FIX L_R x_t+(c_F+δL_FIX L_S)e_t—and simulate it with δ slightly below δ̄: the theorem predicts the error decays as Cρ^t with ρ the spectral radius of M(δ). Any admissible nonlinear instance satisfying Assumptions 1–6 whose error fails to obey the bound (10) at δ<δ̄, or any numerical divergence in that linear system, would refute the claim.

Watch

Extended reading notes

Core claim

The central claim is that convergence of the full two-timescale iteration follows from convergence of its two separated limits, and the separation can be quantified. If δ < c_S(1−c_F)/(L_FIX L_S(c_S+L_R)), the joint error vector measuring distance of the slow state to its target and the fast state to the slow-parameterized fixed-point manifold decays as Cρ^t. The proof reduces one step of the recurrence to a two-by-two comparison matrix whose diagonal entries are the contraction margins and whose off-diagonal entries are the coupling gains; the threshold is exactly the condition that this matrix have spectral radius below one. The same comparison argument covers reduced operators that are on

Load-bearing premise

The load-bearing premise is that the fast map shrinks the distance to its fixed-point manifold by a fixed factor c_F<1 uniformly over all slow states and all points in the state space, with c_F known and global; if contraction is only local, state-dependent, or unquantified, the threshold δ̄ cannot be evaluated and the theorems do not apply.

Editorial extensions

If this is right

  • For any interconnection whose fast map contracts toward a slowly moving manifold and whose reduced map contracts, linear convergence of the joint state is guaranteed as soon as δ is below the explicit threshold; no Lyapunov search is needed.
  • The formula doubles as a tuning rule: smaller coupling gains (L_FIX, L_S, L_R) or larger contraction margins (c_S, 1−c_F) widen the admissible range of δ.
  • Averaged and metrically subregular reduced operators—covering many operator-splitting and optimization iterations—inherit the same timescale separation result, with the contraction constant stated explicitly in terms of the averaging and subregularity parameters.
  • Stochastic versions with expectation-based contraction deliver both convergence in expectation and almost sure convergence, so asynchronous updates, random activations, and packet losses fit the same framework.
  • In feedback optimization, the scheme with plant dynamics as the fast map and gradient descent on the reduced cost as the slow map converges for δ below the bound, whereas δ=1 can diverge even when the reduced method converges.
  • The bound is directly checkable from standard operator constants, which is the contrast with purely existential singular-perturbation results.

Reading between the lines

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

  • The comparison-matrix structure suggests a modular design principle: one can assemble a convergent two-timescale algorithm by certifying the fast component and the reduced component separately, then computing the safety margin; the constants compose exactly as matrix entries.
  • Because δ̄ shrinks as L_FIX or L_S grows, the framework points to a quantitative trade-off between coupling sensitivity and admissible update speed, which could guide sensor, actuator, and controller design in feedback systems.
  • A natural testable extension is to relax the global uniform contraction of the fast map to local or state-dependent rates and ask whether a state-dependent δ(x) can recover linear convergence; the current threshold requires the global constant c_F.
  • For stochastic operators, the expectation-based contraction is weaker than pathwise contraction, so one could explore whether the result persists under Markov-dependent rather than i.i.d. randomness, since the proof only uses the expected comparison inequality.
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

3 major / 3 minor

Summary. The paper studies discrete-time interconnected systems of the form x_{t+1}=x_t+δS(x_t,z_t), z_{t+1}=F(x_t,z_t), viewed as a slow/fast operator interconnection. It shows that if the fast operator contracts uniformly (with rate c_F) toward a slow-state-dependent fixed-point set Z_FIX(x), and the reduced operator x+R(x) is either contractive (Theorem 1) or averaged and metrically subregular (Theorem 2), then for δ below an explicit threshold δ̄ the full interconnection converges linearly to the fixed-point set. The results are extended to stochastic operators with i.i.d. noise (Theorem 3, Corollary 1), yielding convergence in expectation and almost surely. The theory is applied to deterministic and stochastic feedback optimization, producing explicit tuning rules for the timescale parameter. Appendices contain complete proofs of all main results.

Significance. If the results hold, the paper provides a useful operator-theoretic counterpart to continuous-time singular perturbation theory for discrete-time fixed-point iterations. The explicit δ̄ formulas (9), (13), (23) in terms of operator constants are a genuine strength, as is the M-matrix/Schur comparison argument, which avoids Lyapunov-function searches. The proofs are self-contained, and the extension to averaged/subregular and stochastic reduced operators broadens the range of algorithms covered. The feedback-optimization application gives concrete, non-asymptotic tuning rules. However, the advertised coverage of 'paracontractive' fast operators is broader than the actual uniform-contraction assumption, and the 'readily checkable' claim should be tempered for general nonlinear plants.

major comments (3)
  1. [Section II, Assumption 3 (Eq. 5); Section I] Assumption 3 requires a global, uniform contraction rate c_F∈(0,1) toward the fixed-point manifold, d_F(F(x,z),Z_FIX(x))≤c_F d_F(z,Z_FIX(x)). The text labels this 'paracontractive' and states that paracontractivity is weaker than contractivity. Standard paracontraction (strict decrease of distance to each fixed point) does not imply (5). For example, on X×[-1,1] with F(x,z)=z−z^3 and Z_FIX(x)={0}, |F(z)|<|z| for z≠0, but |F(z)|/|z|=1−z^2→1 as z→0, so no uniform c_F exists. Thus Theorems 1–3 do not cover general paracontractive fast operators. The introduction's claim that the results hold 'under a paracontraction property of the fast operator' is inaccurate and should be revised to 'uniform contraction' or accompanied by verifiable sufficient conditions for the uniform bound.
  2. [Section VI-C, Eqs. (42), (44); Conclusions] The paper's central claim that the bounds are 'readily checkable' is only demonstrated on linear plants with known contraction rates (z_{t+1}=z_t+κ(u_t−z_t) and z_{t+1}=z_t+r_t(u_t−z_t)). For a general nonlinear plant satisfying Assumptions 2–5, the constants c_F, L_FIX, L_S, L_R are global and often state-dependent; no procedure, example, or discussion is provided for obtaining them. This does not invalidate the theorems, but it materially narrows the practical scope of the advertised checkability. The manuscript should either provide a nonlinear example with computable constants or add a remark clarifying that the constants must be known a priori and may be difficult to estimate in practice.
  3. [Section V, Assumption 9 (Eq. 16)] The stochastic framework assumes the same fixed-point manifold Z_FIX(x) for every realization r of the noise, i.e., all stochastic fast operators share a common equilibrium manifold. This is a strong requirement: in many stochastic operator interconnections (random coordinate updates, packet drops, asynchronous activations), the fixed-point set depends on the realization. The application in Proposition 2 satisfies the assumption by construction, but the paper does not discuss this restriction or relate it to existing stochastic-approximation settings. Since the stochastic results are advertised as covering 'random updates, asynchronous activations, or packet losses,' this limitation should be explicitly acknowledged.
minor comments (3)
  1. [Section V, Assumption 13 (21b)] The quantification 'for all x,x'∈X and r∈D' is unclear in (21b), where an expectation over r appears. Please state explicitly that (21a) holds for every r, while (21b) is an averaged condition over the distribution of r.
  2. [Theorems 2 and 3, Appendix B] There are several LaTeX formatting artifacts, e.g., '¯x x' instead of a projection notation, and inconsistent use of d_F(z,¯z_x) versus d_F(z,Z_FIX(x)) in Appendix A. These should be cleaned up for readability.
  3. [Appendix E (proof of Theorem 3)] The proof writes E[·] without conditioning on the current state. Since x_t,z_t are random at later steps, the one-step bound should be stated as a conditional expectation given the history, and the final inequality (24) obtained by taking total expectation. The argument is correct but the presentation should be made precise.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: Theorems follow from stated assumptions via explicit comparison inequalities; δ̄ is derived, not fitted, and self-citations are not load-bearing.

full rationale

The derivation chain is self-contained. Theorems 1–3 are proved from Assumptions 1–13 by explicit one-step comparison inequalities (Eqs. 45–61 in the appendices): the error vector is bounded by M(δ) times itself, and δ̄ in Eqs. (9), (13), and (23) is computed from the condition that I2 − M(δ) be a nonsingular M-matrix, using the external characterization [46]. No parameter is fitted and then renamed a prediction: the bound δ̄ is derived from the assumed operator constants, and the simulations evaluate δ̄ analytically and use δ = 0.999 δ̄ without tuning. The self-citations ([10], [43], [45]) appear in the motivation, the preliminary-version reference, and the stochastic feedback-optimization inspiration, but none of these carry a proof step: the convergence proofs invoke only the paper's stated assumptions, standard operator-theoretic facts, and M-matrix theory. The skeptical observation that Assumption 3's uniform contraction with rate c_F is stronger than the paracontraction mentioned in the introduction is a generality/scope issue, not circularity: a theorem conditional on a stronger assumption is not equivalent to its own conclusion. Similarly, the potential difficulty of computing global constants for nonlinear plants concerns practical checkability, not circular construction. Therefore no circular step can be exhibited with a quote and an algebraic reduction.

Assumptions & free parameters 2 free parameters · 6 assumptions · 0 invented entities

The central claim rests on the assumed global operator constants (c_F, c_S, L_FIX, L_S, L_R, α, μ) plus standard operator-theory facts; the paper's contribution is the explicit functional form of δ̄ in (9), (13), (23). Nothing is fitted to data; the residual risk is that the constants must be globally known for the 'readily checkable' claim to hold.

free parameters (2)
  • c_F, c_S, L_FIX, L_S, L_R (operator constants in δ̄)
    Assumptions 2–6 / 9–13 define these as known global inputs; no fitting. But the claimed 'readily checkable' bounds are conditional on them being globally computable, which the paper asserts without giving a procedure for nonlinear plants.
  • α, μ (averagedness / metric-subregularity constants)
    Assumption 7 / 13; determine c_S in Lemma 1/2. In the application derived from the quadratic-growth constant s (α=γL/2, μ=2/(γs)); in general μ is rarely available.
assumptions (6)
  • domain assumption Uniform fast contraction toward Z_FIX(x) (Assumption 3 / 10, Eq. 5 / 18)
    Load-bearing stability of the boundary-layer subsystem; stronger than the 'paracontraction' described in the text.
  • domain assumption Well-posed, Lipschitz-moving fixed-point manifold Z_FIX(x) (Assumption 2 / 9)
    Replaces the classical equilibrium-manifold assumption of singular perturbation theory; requires known L_FIX.
  • domain assumption Reduced operator well-defined and contracting / averaged+subregular (Assumptions 4, 6, 7 / 11, 13)
    Stability of the reduced system, the second half of the classical singular-perturbation assumption pair.
  • standard math M-matrix characterization of Schur nonnegative matrices (Plemmons [46])
    Used to convert the 2×2 comparison inequality into the explicit δ-bound (49).
  • standard math Gradient-descent contraction on strongly convex + smooth / convex + quadratic-growth costs (standard, cited to [21])
    Used to compute c_S and α, μ in Propositions 1–2 from s, L, γ.
  • domain assumption Application: plant z→f(z,u) globally c_F-contractive; ∇z_ss and ∇ℓ globally bounded/Lipschitz; ℓ∘z_ss strongly convex/convex with known s, L
    Application-specific verifiability of the framework's assumptions; rarely available globally for nonlinear plants.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Timescale Separation Through the Lens of Operator Theory." pith.science (2026). https://pith.science/paper/UMNMYKL4

@misc{pith2026260802443,
  author       = {Pith},
  title        = {Pith review of: Timescale Separation Through the Lens of Operator Theory},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/UMNMYKL4}},
  note         = {Machine review of arXiv:2608.02443}
}
read the original abstract

Timescale separation is a powerful tool for analyzing interconnected dynamical systems. Meanwhile, operator theory provides a general framework for studying the convergence of iterative methods formulated as fixed-point iterations, including algorithms arising in optimization, learning, and control. In this paper, we bridge these two areas by establishing timescale separation results for fixed-point iterations induced by both deterministic and stochastic operators. As customary in timescale separation, our results involve auxiliary systems that arise from the original interconnection in the limit as the timescale parameter tends to zero and separately capture the dynamics induced by the slow and fast operators. The proposed operator-theoretic framework yields explicit and readily checkable bounds on this tunable parameter, expressed in terms of standard operator constants. To illustrate the applicability of our results, we employ them to prove the convergence properties of a feedback optimization scheme in both deterministic and stochastic settings.

Figures

Figures reproduced from arXiv: 2608.02443 by the authors.

Figure 1
Figure 1. Graphical representation of system (1). This peculiar class of systems can often be conveniently analyzed through two auxiliary descriptions: a boundary layer subsystem, obtained by freezing the slow variables, and a reduced subsystem, obtained by replacing in the slow dynamics the fast variables with their limiting response. This idea lies at the core of singular perturbation theory and has played a central role in… view at source ↗
Figure 2
Figure 2. Graphical representation of the boundary layer system (characterized in Assumption 3) associated to the interconnected [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Graphical representation of the reduced system (7) associated to the interconnected system (1). [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Error ∥ut − u⋆∥ achieved in the strongly convex case by the closed-loop system (29) with different values of δ, together with the reduced system (34). All three cases are tested with γ = 2/(s + L). To test (31) as well, we consider the same setup as before, but generat…
Figure 5
Figure 5. Figure 5: Optimality error ∥∇ℓ(zss(ut))∥ achieved in the convex case by the closed-loop system (29) with different values of δ, together with the reduced system (34). All three cases are tested with γ = 1/L. Then, we test the stochastic result in Proposition 2. We consider the s…
Figure 6
Figure 6. Figure 6: Error ∥ut − u⋆∥ achieved in the strongly convex case by the closed-loop system (37) with different values of δ, together with the reduced system (34). All three cases are tested with γ = 2/(s + L). 0 50 100 150 200 10−16 10−9 10−2 105 Iteration t ∥∇ℓ(zss(ut))∥ Reduced …
Figure 7
Figure 7. Figure 7: Optimality error ∥∇ℓ(zss(ut))∥ achieved in the convex case by the closed-loop system (37) with different values of δ, together with the reduced system (34). All three cases are tested with γ = 1/L. VII. CONCLUSIONS In this paper, we developed an operator-theoretic fram…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

46 extracted references · 1 linked inside Pith

  1. [43]

    Contraction theory and timescale separation for interconnected nonlinear systems,

    G. Carnevale, N. Bastianello, R. Carli, L. Schenato, and G. Notarstefano, “Contraction theory and timescale separation for interconnected nonlinear systems,” in2026 European Control Conference (ECC). IEEE, 2026

    2026

  2. [1]

    Singular perturbation method for reducing the model order in optimal control design,

    P. Kokotovic and P. Sannuti, “Singular perturbation method for reducing the model order in optimal control design,”IEEE Transactions on Automatic Control, vol. 13, no. 4, pp. 377–384, 1968

    1968

  3. [2]

    Near-optimum design of linear systems by a singular perturbation method,

    P. Sannuti and P. Kokotovic, “Near-optimum design of linear systems by a singular perturbation method,”IEEE Transactions on Automatic Control, vol. 14, no. 1, pp. 15–22, 1969

    1969

  4. [3]

    Singular perturbations and order reduction in control theory—an overview,

    P. V . Kokotovic, R. E. O’Malley Jr, and P. Sannuti, “Singular perturbations and order reduction in control theory—an overview,”Automatica, vol. 12, no. 2, pp. 123–132, 1976

    1976

  5. [4]

    Kokotovi ´c, H

    P. Kokotovi ´c, H. K. Khalil, and J. O’Reilly,Singular perturbation methods in control: analysis and design. SIAM, 1999

    1999

  6. [5]

    Multi-time scale control and optimization via averaging and singular perturbation theory: From ODEs to hybrid dynamical systems,

    M. Abdelgalil, D. E. Ochoa, and J. I. Poveda, “Multi-time scale control and optimization via averaging and singular perturbation theory: From ODEs to hybrid dynamical systems,”Annual Reviews in Control, vol. 56, p. 100926, 2023

    2023

  7. [6]

    Averaging analysis for discrete time and sampled data adaptive systems,

    E.-W. Bai, L.-C. Fu, and S. S. Sastry, “Averaging analysis for discrete time and sampled data adaptive systems,”IEEE Transactions on Circuits and Systems, vol. 35, no. 2, pp. 137–148, 1988

    1988

  8. [7]

    On analysis of discrete singularly perturbed non-linear systems: application to the study of stability properties,

    R. Bouyekhf and A. El Moudni, “On analysis of discrete singularly perturbed non-linear systems: application to the study of stability properties,”Journal of the Franklin Institute, vol. 334, no. 2, pp. 199–212, 1997

    1997

Show all 46 references
  1. [8]

    A unified framework for input-to-state stability in systems with two time scales,

    A. R. Teel, L. Moreau, and D. Nesic, “A unified framework for input-to-state stability in systems with two time scales,”IEEE Transactions on Automatic Control, vol. 48, no. 9, pp. 1526–1544, 2003

    2003

  2. [9]

    Discrete singularly perturbed control problems (a survey),

    G. Kurina, M. Dmitriev, and D. S. Naidu, “Discrete singularly perturbed control problems (a survey),”Dyn. Contin. Discrete Impuls. Syst. Ser. B: Appl. Algorithms, vol. 24, pp. 335–370, 2017

    2017

  3. [10]

    Timescale separation for discrete-time nonlinear stochastic systems,

    G. Carnevale and G. Notarstefano, “Timescale separation for discrete-time nonlinear stochastic systems,”IEEE Control Systems Letters, vol. 8, pp. 2133–2138, 2024

    2024

  4. [11]

    On contraction analysis for non-linear systems,

    W. Lohmiller and J.-J. E. Slotine, “On contraction analysis for non-linear systems,”Automatica, vol. 34, no. 6, pp. 683–696, 1998

    1998

  5. [12]

    Contraction methods for nonlinear systems: A brief introduction and some open problems,

    Z. Aminzare and E. D. Sontag, “Contraction methods for nonlinear systems: A brief introduction and some open problems,” in53rd IEEE Conference on Decision and Control. IEEE, 2014, pp. 3835–3847

    2014

  6. [13]

    An operator splitting approach for distributed generalized Nash equilibria computation,

    P. Yi and L. Pavel, “An operator splitting approach for distributed generalized Nash equilibria computation,”Automatica, vol. 102, pp. 111–121, 2019. 17

    2019

  7. [14]

    Semi-decentralized generalized Nash equilibrium seeking in monotone aggregative games,

    G. Belgioioso and S. Grammatico, “Semi-decentralized generalized Nash equilibrium seeking in monotone aggregative games,”IEEE Transactions on Automatic Control, 2021

    2021

  8. [15]

    Stochastic generalized Nash equilibrium-seeking in merely monotone games,

    B. Franci and S. Grammatico, “Stochastic generalized Nash equilibrium-seeking in merely monotone games,”IEEE Transactions on Automatic Control, vol. 67, no. 8, pp. 3905–3919, 2021

    2021

  9. [16]

    Distributed generalized Nash equilibrium seeking: An operator-theoretic perspective,

    G. Belgioioso, P. Yi, S. Grammatico, and L. Pavel, “Distributed generalized Nash equilibrium seeking: An operator-theoretic perspective,”IEEE Control Systems Magazine, vol. 42, no. 4, pp. 87–102, 2022

    2022

  10. [17]

    Synchronization in networks of nonlinear systems: Contraction analysis via Riemannian metrics and deep-learning for feedback estimation,

    M. Giaccagli, S. Zoboli, D. Astolfi, V . Andrieu, and G. Casadei, “Synchronization in networks of nonlinear systems: Contraction analysis via Riemannian metrics and deep-learning for feedback estimation,”IEEE Transactions on Automatic Control, vol. 69, no. 11, pp. 8041–8048, 2024

    2024

  11. [18]

    Perspectives on contractivity in control, optimization, and learning,

    A. Davydov and F. Bullo, “Perspectives on contractivity in control, optimization, and learning,”IEEE Control Systems Letters, vol. 8, pp. 2087–2098, 2024

  12. [19]

    The END: Estimation network design for games under partial-decision information,

    M. Bianchi and S. Grammatico, “The END: Estimation network design for games under partial-decision information,”IEEE Transactions on Control of Network Systems, vol. 11, no. 4, pp. 2200–2212, 2024

    2024

  13. [20]

    Finite-sample guarantees for data-driven forward-backward operator methods,

    F. Fabiani and B. Franci, “Finite-sample guarantees for data-driven forward-backward operator methods,”IEEE Transactions on Automatic Control, 2026

    2026

  14. [21]

    Primer on monotone operator methods,

    E. K. Ryu and S. Boyd, “Primer on monotone operator methods,”Appl. comput. math, vol. 15, no. 1, pp. 3–43, 2016

    2016

  15. [22]

    H. H. Bauschke and P. L. Combettes,Convex Analysis and Monotone Operator Theory in Hilbert Spaces. Springer, 2017

    2017

  16. [23]

    Contraction and robustness of continuous time primal-dual dynamics,

    H. D. Nguyen, T. L. Vu, K. Turitsyn, and J.-J. Slotine, “Contraction and robustness of continuous time primal-dual dynamics,”IEEE Control Systems letters, vol. 2, no. 4, pp. 755–760, 2018

    2018

  17. [24]

    SuperMann: a superlinearly convergent algorithm for finding fixed points of nonexpansive operators,

    A. Themelis and P. Patrinos, “SuperMann: a superlinearly convergent algorithm for finding fixed points of nonexpansive operators,”IEEE Transactions on Automatic Control, vol. 64, no. 12, pp. 4875–4890, 2019

    2019

  18. [25]

    Multi-agent optimization and learning: A non-expansive operators perspective,

    N. Bastianello, L. Schenato, and R. Carli, “Multi-agent optimization and learning: A non-expansive operators perspective,” inEncyclopedia of Systems and Control Engineering, 1st ed., Z. Ding, Ed. Elsevier, 2026, pp. 630–638

    2026

  19. [26]

    On weakly contracting dynamics for convex optimization,

    V . Centorrino, A. Davydov, A. Gokhale, G. Russo, and F. Bullo, “On weakly contracting dynamics for convex optimization,”IEEE Control Systems Letters, vol. 8, pp. 1745–1750, 2024

    2024

  20. [27]

    Non-Euclidean monotone operator theory and applications,

    A. Davydov, S. Jafarpour, A. V . Proskurnikov, and F. Bullo, “Non-Euclidean monotone operator theory and applications,”Journal of Machine Learning Research, vol. 25, no. 307, pp. 1–33, 2024

    2024

  21. [28]

    Time-varying convex optimization: A contraction and equilibrium tracking approach,

    A. Davydov, V . Centorrino, A. Gokhale, G. Russo, and F. Bullo, “Time-varying convex optimization: A contraction and equilibrium tracking approach,” IEEE Transactions on Automatic Control, 2025

    2025

  22. [29]

    Online optimization as a feedback controller: Stability and tracking,

    M. Colombino, E. Dall’Anese, and A. Bernstein, “Online optimization as a feedback controller: Stability and tracking,”IEEE Transactions on Control of Network Systems, vol. 7, no. 1, pp. 422–432, 2019

    2019

  23. [30]

    Timescale separation in autonomous optimization,

    A. Hauswirth, S. Bolognani, G. Hug, and F. D ¨orfler, “Timescale separation in autonomous optimization,”IEEE Transactions on Automatic Control, vol. 66, no. 2, pp. 611–624, 2020

    2020

  24. [31]

    Optimization algorithms as robust feedback controllers,

    A. Hauswirth, Z. He, S. Bolognani, G. Hug, and F. D ¨orfler, “Optimization algorithms as robust feedback controllers,”Annual Reviews in Control, vol. 57, p. 100941, 2024

    2024

  25. [32]

    Nonconvex distributed feedback optimization for aggregative cooperative robotics,

    G. Carnevale, N. Mimmo, and G. Notarstefano, “Nonconvex distributed feedback optimization for aggregative cooperative robotics,”Automatica, vol. 167, p. 111767, 2024

    2024

  26. [33]

    A Lyapunov function for the combined system-optimizer dynamics in inexact model predictive control,

    A. Zanelli, Q. Tran-Dinh, and M. Diehl, “A Lyapunov function for the combined system-optimizer dynamics in inexact model predictive control,” Automatica, vol. 134, p. 109901, 2021

    2021

  27. [34]

    Sampled-data systems: Stability, contractivity and single-iteration suboptimal MPC,

    Y . Chen, F. Bullo, and E. Dall’Anese, “Sampled-data systems: Stability, contractivity and single-iteration suboptimal MPC,”IEEE Transactions on Automatic Control, 2026

    2026

  28. [35]

    Suboptimal and reduced-order MPC via timescale separation,

    S. Di Gregorio, G. Carnevale, and G. Notarstefano, “Suboptimal and reduced-order MPC via timescale separation,”arXiv preprint arXiv:2606.24641, 2026

    2026 arXiv

  29. [36]

    ADMM-tracking gradient for distributed optimization over asynchronous and unreliable networks,

    G. Carnevale, N. Bastianello, G. Notarstefano, and R. Carli, “ADMM-tracking gradient for distributed optimization over asynchronous and unreliable networks,”IEEE Transactions on Automatic Control, vol. 70, no. 8, pp. 5160–5175, 2025

    2025

  30. [37]

    Tracking-based distributed equilibrium seeking for aggregative games,

    G. Carnevale, F. Fabiani, F. Fele, K. Margellos, and G. Notarstefano, “Tracking-based distributed equilibrium seeking for aggregative games,”IEEE Transactions on Automatic Control, vol. 69, no. 9, pp. 6026–6041, 2024

    2024

  31. [38]

    A unifying system theory framework for distributed optimization and games,

    G. Carnevale, N. Mimmo, and G. Notarstefano, “A unifying system theory framework for distributed optimization and games,”IEEE Transactions on Automatic Control, vol. 70, no. 11, pp. 7318–7333, 2025

    2025

  32. [39]

    Asynchronous distributed optimization over lossy networks via relaxed ADMM: Stability and linear convergence,

    N. Bastianello, R. Carli, L. Schenato, and M. Todescato, “Asynchronous distributed optimization over lossy networks via relaxed ADMM: Stability and linear convergence,”IEEE Transactions on Automatic Control, vol. 66, no. 6, pp. 2620–2635, 2020

    2020

  33. [40]

    ADMM for dynamic average consensus over imperfect networks,

    N. Bastianello and R. Carli, “ADMM for dynamic average consensus over imperfect networks,”IFAC-PapersOnLine, vol. 55, no. 13, pp. 228–233, 2022

    2022

  34. [41]

    Modular distributed nonconvex learning with error feedback,

    G. Carnevale and N. Bastianello, “Modular distributed nonconvex learning with error feedback,”IEEE Control Systems Letters, 2025

    2025

  35. [42]

    Online feedback optimization and singular perturbation via contraction theory,

    L. Cothren, F. Bullo, and E. Dall’Anese, “Online feedback optimization and singular perturbation via contraction theory,”SIAM Journal on Control and Optimization, vol. 64, no. 3, pp. 1482–1506, 2026

    2026

  36. [44]

    Optimization and learning in open multi-agent systems,

    D. Deplano, N. Bastianello, M. Franceschelli, and K. H. Johansson, “Optimization and learning in open multi-agent systems,”IEEE Transactions on Automatic Control, 2026

    2026

  37. [45]

    Almost sure convergence in feedback optimization via stochastic timescale separation,

    G. Carnevale and G. Notarstefano, “Almost sure convergence in feedback optimization via stochastic timescale separation,” in2025 IEEE 64th Conference on Decision and Control (CDC). IEEE, 2025, pp. 585–590

    2025

  38. [46]

    M-matrix characterizations. I—nonsingular M-matrices,

    R. J. Plemmons, “M-matrix characterizations. I—nonsingular M-matrices,”Linear Algebra and its applications, vol. 18, no. 2, pp. 175–188, 1977

    1977

Pith tools

Reviewed August 4, 2026 · model on record in the stance chip above.