Proximal gradient flow and Douglas-Rachford splitting dynamics: global exponential stability via integral quadratic constraints

Mihailo R. Jovanovi\'c; Sepideh Hassan-Moghaddam

Proximal gradient flow and Douglas-Rachford splitting dynamics: global exponential stability via integral quadratic constraints

T0 review · 0 major / 5 minor · reviewed 2026-08-14 · deepseek-v4-flash

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

Pith's one-line read The paper proves global exponential stability of the continuous-time proximal gradient and Douglas-Rachford splitting flows, with explicit rates determined by the strong-convexity and smoothness constants.

desk verdict A sound IQC-based analysis with a genuinely new continuous-time DR flow result; fix the appendix sign error and Section 4.2 before publication. read the letter →

arxiv 1908.09043 v2 pith:AD2FJB77 submitted 2019-08-23 math.OC cs.LGcs.SYeess.SYmath.DS

Sepideh Hassan-Moghaddam , Mihailo R. Jovanović This is my paper · ORCID

classification math.OCcs.LGcs.SYeess.SYmath.DS MSC 90C2565K1093D23

keywords proximalgradientflowDouglas-Rachfordsplittingglobalexponentialstabilityintegralquadraticconstraintsforward-backwardenvelopePLconditioncompositeoptimizationnonsmooth

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

This paper treats proximal gradient and Douglas-Rachford splitting as ordinary differential equations and asks whether their equilibria are globally exponentially stable. For composite objectives in which the smooth part is strongly convex with Lipschitz gradient and the nonsmooth part is convex, the answer is yes for every step size $\mu \in (0, 2/L_f)$, with an explicit rate $\rho \le 1 - \max\{|1-\mu m_f|, |1-\mu L_f|\}$. The proof rewrites each algorithm as a feedback loop between a stable linear system and a nonlinear proximal step that is a strict contraction, then uses integral quadratic constraints, a control-theoretic stability certificate, to certify global exponential stability. Under a proximal gradient-dominance condition, the same ideas show the forward-backward envelope, a smooth surrogate for the objective, converges exponentially even without strong convexity, although the trajectory itself need not. The payoff is a unified explanation of convergence whose rates are tied directly to the conditioning of the problem.

What carries the argument

The load-bearing object is a pointwise quadratic inequality for the nonlinear map $u(\xi) = \mathrm{prox}_{\mu g}(\xi - \mu\nabla f(\xi))$: for any two points, $\|u(\xi)-u(\hat{\xi})\|^2 \le \sigma^2 \|\xi-\hat{\xi}\|^2$ with $\sigma = \max\{|1-\mu m_f|, |1-\mu L_f|\}$. This sector bound is what integral quadratic constraints require, and it makes $u$ a strict contraction exactly when $\mu < 2/L_f$. It is derived by combining firm nonexpansiveness of the proximal operator with a two-sided inequality that controls the inner product of two gradient differences in terms of both the distance between points and the norm of the gradient difference. The same bound, with the reflected proximal operator $R_{\mu f} = 2\,\mathrm{prox}_{\mu f} - I$ in place of $u$, covers the Douglas-Rachford nonlinearity $R_{\mu g}R_{\mu f}$. Feeding this bound into the exponential-stability test turns the global stability question into the solvability of a small matrix inequality, which holds precisely when $\rho \le 1-\sigma$.

What would settle it

Take the scalar problem $f(x)=\frac{L_f}{2}x^2$ with $g=0$; the proximal gradient flow becomes the linear ODE $\dot{x}=-\mu L_f x$, whose exact decay exponent is $\mu L_f$, while the theorem certifies the rate $1-\sigma$ with $\sigma=|1-\mu L_f|$, so checking $\mu$ across $(0,2/L_f)$ settles whether the claimed rate ever overstates the true contraction.

Watch

Extended reading notes

Core claim

The central discovery is that the continuous-time proximal gradient flow, $\dot{x} = -(x - \mathrm{prox}_{\mu g}(x - \mu \nabla f(x)))$, and the Douglas-Rachford splitting flow, $\dot{z} = -z + R_{\mu g}R_{\mu f}(z)$, are globally exponentially stable when $f$ is $m_f$-strongly convex with $L_f$-Lipschitz gradient, $g$ is a convex nonsmooth function, and $\mu \in (0, 2/L_f)$. The exponential rate is $\rho \le 1 - \sigma$ with $\sigma = \max\{|1-\mu m_f|, |1-\mu L_f|\}$, and the choice $\mu = 2/(L_f+m_f)$ gives $\rho \le 2/(\kappa+1)$, where $\kappa = L_f/m_f$ is the condition number. The proof is uniform: both nonlinear maps satisfy the same pointwise quadratic inequality, so both fit the same stability test. When strong convexity is absent but the proximal gradient-dominance condition holds, the paper shows the forward-backward envelope $F_\mu(x(t))$ decays as $F_\mu(x(t)) - F_\mu^\star \le e^{-\gamma\mu(1-\mu L_f)t}(F_\mu(x(0)) - F_\mu^\star)$.

Load-bearing premise

The proof needs the proximal step's nonlinear map to be a strict contraction, which forces the smooth part to be strongly convex with Lipschitz gradient and the step size to satisfy $\mu < 2/L_f$; once $\sigma = \max\{|1-\mu m_f|, |1-\mu L_f|\}$ reaches 1, the algebra behind the stability certificate stops working.

Editorial extensions

If this is right

The standard proximal gradient and Douglas-Rachford iterations are explicit forward-Euler discretizations of these flows, so the continuous-time theorem gives an idealized convergence rate for the algorithms they discretize.
For a fixed condition number $\kappa = L_f/m_f$, the best step size in this analysis is $\mu = 2/(L_f+m_f)$, yielding the rate $2/(\kappa+1)$, the same condition-number dependence as classical gradient descent in the smooth case.
The same IQC proof applies to the Douglas-Rachford flow formulated on the dual problem, covering the ADMM-equivalent algorithm whenever the constraint matrix has full row rank.
Without strong convexity, the proximal gradient-dominance condition still forces the forward-backward envelope to converge exponentially at rate $\gamma\mu(1-\mu L_f)$, but it does not give an exponential rate for the distance to the optimizer.
Because both algorithms share one quadratic characterization, any tightening of the sector bound would automatically improve the certified rate for both flows.

Reading between the lines

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

The rate bound is designed for the continuous-time dynamics; the discrete algorithms' actual rates should differ from $1-\sigma$ by an extra discretization factor, and setting up a matched stability test for the discrete iteration could quantify that gap.
The proof uses only the strong-convexity and smoothness constants of $f$, so any nonlinear map with the same contraction parameters would satisfy the same stability certificate; this suggests immediate extensions to preconditioned or inexact proximal steps.
Inside the proximal gradient-dominance branch, the rate $\gamma\mu(1-\mu L_f)$ is maximized at $\mu = 1/(2L_f)$, giving $\gamma/(4L_f)$; the paper does not discuss this optimization, but it follows directly from the displayed bound.
Since the Douglas-Rachford flow is stable for a merely convex nonsmooth part, the continuous-time viewpoint may help analyze alternating schemes that lack an obvious Lyapunov function.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Desk Editor's Note

read the letter

Colleague,

The short version: the main theorems check out, and the continuous-time Douglas-Rachford splitting flow is a real new contribution. But the paper has a couple of warts—an unproved claim in Section 4.2 and a sign error in Appendix A—that need a pass before I'd sign off.

What's actually new: the DR splitting flow (26) and its global exponential stability theorem (Theorem 6). The proof is neat: R_μg is firmly nonexpansive, R_μf is σ-contractive with the same σ as the proximal gradient nonlinearity, so the same IQC multiplier applies and the LMI (22) gives the rate. That is a clean observation. The PL-based envelope convergence theorem (Theorem 3) is also new, and it's a straightforward Lyapunov argument that works. The main proofs are complete and the rates are explicit; the optimal μ = 2/(L_f+m_f) recovers the expected 2/(κ+1) rate. No fitted parameters, no circularity.

Soft spots, in proportion. The novelty claim in the introduction is a bit strong: the proximal gradient flow already appeared in the authors' own CDC 2018 paper [27], so only the DR flow is genuinely first here. Section 4.2 says stability of DR on the dual problem is 'readily established' and provides no proof; that's an assertion, not a result, and it should be either proven or trimmed. Appendix A has a sign error in (A.4): the term multiplying ⟨∇f, G⟩ should be +(μκ−1), not −(μκ−1). It's a typo—the reader and the stress-test both caught it, and with the sign fixed the derivation works—but it needs flagging. These are all minor issues.

Bottom line: the central stability claims hold up. The paper is honest about its limitations, including the absence of an x(t) rate under PL without strong convexity. It is written for people working on continuous-time optimization dynamics and IQC-based convergence. I'd send it to review; the correct verdict is minor revision, not accept as is.

Best,

Referee Report

0 major / 5 minor

Summary. The paper introduces continuous-time models for proximal gradient and Douglas-Rachford splitting applied to nonsmooth composite problems and analyzes their global stability. For problems of the form min_x f(x)+g(x) with f strongly convex and ∇f Lipschitz, Theorem 2 proves that the equilibrium of the proximal gradient flow (15) is globally exponentially stable for μ∈(0,2/L_f) with rate ρ≤1−σ, where σ=max{|1−μm_f|,|1−μL_f|}. Lemma 1 and Lemma 4 establish the underlying contraction and sector properties, and Theorem 6 carries the same rate to the DR splitting flow (26). Under a proximal Polyak-Lojasiewicz condition (Assumption 2), Theorem 3 establishes exponential decay of the forward-backward envelope. The paper also discusses extensions to dual DR dynamics and connects the proximal PL condition to a known PL formulation in Appendix A.

Significance. If the results hold, the paper provides a clean, control-theoretic treatment of continuous-time proximal algorithms with explicit exponential rates. The central proofs are complete: Lemma 1's sector bound (17a) follows from firm nonexpansiveness of the proximal operator, strong convexity, and Nesterov's inequality; the LMI (22)–(23) in Theorem 2 is verified with P=pI; and Theorem 3 is a standard Lyapunov argument. The paper is also honest about the limitation of the PL branch in Remark 4. The rates are explicit and the assumptions are standard, and the proofs do not rely on fitted constants. These are useful tools for the optimization-as-dynamical-systems literature and the paper is likely to be of interest to readers of Automatica.

minor comments (5)

[Section 4.2] The global exponential stability claim for the dual DR dynamics (30) is asserted as 'readily established' without proof; because the conjugates f1 and g1 require verifying the relevant strong-convexity, smoothness, and IQC conditions, please either provide the proof or explicitly mark this as a conjecture.
[Appendix A, Eq. (A.4)] There is a sign error in the term multiplying ⟨∇f(x), G_μ(x)⟩: it should be +(μκ−1) rather than −(μκ−1). As written, the transition from (A.4) to the displayed bound with γ=2κ/|μκ−1| is not valid. Because Theorem 3 assumes (8) directly, this error does not undermine the main theorem, but the appendix should be corrected and the derivation checked.
[Lemma 5] In the proof of Lemma 5, the expression 'prox_{μf}' should read 'prox_{μg}'; otherwise the displayed identity concerns the wrong resolvent.
[Theorem 6, proof] In the proof of Theorem 6, 'systems (11)' should be 'systems (16)'; the feedback interconnection used there is (16a) with nonlinearity (27).
[Remark 2] The sentence 'this is the best achievable convergence rate for system (15)' is a strong optimality claim with no proof or citation; please add a derivation or soften the statement.

Circularity Check

0 steps flagged · score 1.0 of 10

No load-bearing circularity found: the exponential-stability proofs derive their rates from the stated assumptions plus external IQC and convex-analysis tools, while the authors' self-citations are contextual rather than essential.

full rationale

The central derivations are self-contained with respect to their assumptions. Lemma 1's sector inequality (17a) is obtained by combining firm nonexpansiveness of the proximal map [6], strong convexity of f, Lf-Lipschitz continuity of ∇f, and Nesterov's inequality (6) from [40]; this yields sigma = max{|1−μmf|, |1−μLf|} by computation, not by definition of the conclusion. Theorem 2 then applies the external IQC decay-rate condition of [42, Theorem 3] to the resulting sector bound, solving the 1x1 LMI (22)-(23) to obtain rho ≤ 1−sigma. Theorem 6 uses the same sector characterization because Rμg is firmly nonexpansive and Rμf is sigma-contractive, a property proved in Lemmas 4 and 5. Theorem 3 is a direct Lyapunov argument under the proximal PL inequality (8), and the exponential rate γμ(1−μLf) is explicitly implied by that assumed inequality. No fitted constant is relabeled as a prediction, and no result in the paper is equivalent to its own input by construction. The self-citations to the authors' earlier work [11] and [27] are used for background, for the proximal augmented Lagrangian representation, and for related framing; they are not the load-bearing evidence for Theorems 2, 3, or 6. Two non-circular completeness issues should be noted: Section 4.2 asserts that dual Douglas-Rachford stability is 'readily established' without providing the proof, and Appendix A contains an apparent sign error near (A.4) in deriving the proximal PL bound. These are correctness or exposition concerns, not circular reductions, and they do not affect the main stability theorems.

Assumptions & free parameters 1 free parameters · 6 assumptions · 0 invented entities

The central proofs depend on standard convex-analysis facts and the two stated regularity assumptions; no unknown constants are fitted and no new entities are postulated.

free parameters (1)

Proximal step size mu
Algorithm parameter in the flows (15) and (26); the theorems require mu in (0,2/L_f) (and mu in (0,1/L_f) for the PL theorem). The rates depend on mu but mu is not fitted to data.

assumptions (6)

domain assumption f is m_f-strongly convex and gradient L_f-Lipschitz (Assumption 1)
Main regularity condition for Theorems 2 and 6; without it the sector constant sigma < 1 cannot be guaranteed.
standard math prox_{μg} is firmly nonexpansive and Moreau envelope is continuously differentiable (Parikh-Boyd [6])
Used in Lemma 1 and throughout to characterize the nonlinear block u(x) = prox_{μg}(x - mu grad f(x)).
standard math Nesterov inequality (6) for strongly convex smooth functions
Used in Lemma 1 to derive the contraction bound (20).
standard math Hu-Seiler IQC exponential stability criterion [42, Theorem 3]
Converts sector inequality (17a) and LMI (21) into the global exponential decay estimate; not reproved.
domain assumption f twice continuously differentiable with Hessian bounded above by L_f I, and proximal PL inequality (8) (Assumption 2)
Supplies the exponential rate in Theorem 3 via the Lyapunov function V = F_mu - F*_mu.
standard math Dual equivalence between DR splitting and ADMM (Eckstein-Bertsekas [45], Gabay [46])
Used in Section 4.2 to discuss DR splitting on dual problems.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Proximal gradient flow and Douglas-Rachford splitting dynamics: global exponential stability via integral quadratic constraints." pith.science (2026). https://pith.science/paper/AD2FJB77

@misc{pith2026190809043,
  author       = {Pith},
  title        = {Pith review of: Proximal gradient flow and Douglas-Rachford splitting dynamics: global exponential stability via integral quadratic constraints},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/AD2FJB77}},
  note         = {Machine review of arXiv:1908.09043}
}

read the original abstract

Many large-scale and distributed optimization problems can be brought into a composite form in which the objective function is given by the sum of a smooth term and a nonsmooth regularizer. Such problems can be solved via a proximal gradient method and its variants, thereby generalizing gradient descent to a nonsmooth setup. In this paper, we view proximal algorithms as dynamical systems and leverage techniques from control theory to study their global properties. In particular, for problems with strongly convex objective functions, we utilize the theory of integral quadratic constraints to prove the global exponential stability of the equilibrium points of the differential equations that govern the evolution of proximal gradient and Douglas-Rachford splitting flows. In our analysis, we use the fact that these algorithms can be interpreted as variable-metric gradient methods on the suitable envelopes and exploit structural properties of the nonlinear terms that arise from the gradient of the smooth part of the objective function and the proximal operator associated with the nonsmooth regularizer. We also demonstrate that these envelopes can be obtained from the augmented Lagrangian associated with the original nonsmooth problem and establish conditions for global exponential convergence even in the absence of strong convexity.

Figures

Figures reproduced from arXiv: 1908.09043 by the authors.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

46 extracted references · 42 canonical work pages

[27]

On the exponential convergence rate of proximal gradient ﬂow algorithms,

S. Hassan-Moghaddam and M. R. Jovanovi´ c, “On the exponential convergence rate of proximal gradient ﬂow algorithms,” in Proceedings of the 57th IEEE Conference on Decision and Control , Miami, FL, 2018, pp. 4246–4251

work page 2018
[1]

Distributed subgradient met hods for multiagent optimization,

A. Nedi´ c and A. Ozdaglar, “Distributed subgradient met hods for multiagent optimization,” IEEE Trans. on Automat. Control, vol. 54, no. 1, pp. 48–61, 2009

work page 2009
[2]

A control perspective for centraliz ed and distributed convex optimization,

J. W ang and N. Elia, “A control perspective for centraliz ed and distributed convex optimization,” in Proceedings of the 50th IEEE Conference on Decision and Control , 2011, pp. 3800–3805

work page 2011
[3]

New primal- dual proximal algorithm for distributed optimization,

P. Latafat, L. Stella, and P. Patrinos, “New primal- dual proximal algorithm for distributed optimization,” in Proceedings of the 55th IEEE Conference on Decision and Control, 2016, pp. 1959–1964

work page 2016
[4]

A new randomized block-coordinate primal-dual proximal algori thm for distributed optimization,

P. Latafat, N. Freris, and P. Patrinos, “A new randomized block-coordinate primal-dual proximal algori thm for distributed optimization,” IEEE Trans. Automat. Control, 2019, doi:10.1109/TAC.2019.2906924

arXiv 2019
[5]

A fast iterative shrinkage- thresholding algorithm for linear inverse problems,

A. Beck and M. Teboulle, “A fast iterative shrinkage- thresholding algorithm for linear inverse problems,” SIAM J. Imaging Sci. , vol. 2, no. 1, pp. 183–202, 2009

work page 2009
[6]

Proximal algorithms,

N. Parikh and S. Boyd, “Proximal algorithms,” Found. Trends Opt., vol. 1, no. 3, pp. 123–231, 2013

work page 2013
[7]

Studies in linear and non-linear programming,

K. J. Arrow, L. Hurwicz, and H. Uzawa, “Studies in linear and non-linear programming,” 1958

work page 1958

Show all 46 references

[8]

Stability of primal–dual gra dient dynamics and applications to network optimization,

D. Feijer and F. Paganini, “Stability of primal–dual gra dient dynamics and applications to network optimization,” Automatica, vol. 46, no. 12, pp. 1974–1981, 2010

1974
[9]

Asymptotic convergence of constrained primal–dual dynamics,

A. Cherukuri, E. Mallada, and J. Cort´ es, “Asymptotic convergence of constrained primal–dual dynamics,” Syst. Control Lett., vol. 87, pp. 10–15, 2016

2016
[10]

The rol e of convexity on saddle-point dynamics: Lyapunov function and robustness,

A. Cherukuri, E. Mallada, S. Low, and J. Cortes, “The rol e of convexity on saddle-point dynamics: Lyapunov function and robustness,” IEEE Trans. Automat. Control , vol. 63, no. 8, pp. 2449–2464, 2018

2018
[11]

The proximal augmented Lagrangian method for nonsmooth composite optimization,

N. K. Dhingra, S. Z. Khong, and M. R. Jovanovi´ c, “The proximal augmented Lagrangian method for nonsmooth composite optimization,” IEEE Trans. Automat. Control , vol. 64, no. 7, pp. 2861–2868, July 2019

2019
[12]

On the exponential stability of primal- dual gradient dynamics,

G. Qu and N. Li, “On the exponential stability of primal- dual gradient dynamics,” IEEE Control Syst. Lett. , vol. 3, no. 1, pp. 43–48, 2018

2018
[13]

A diﬀerential equation for modeling Nesterov’s accelerated gradient method: Theo ry and insights,

W. Su, S. Boyd, and E. Candes, “A diﬀerential equation for modeling Nesterov’s accelerated gradient method: Theo ry and insights,” J. Mach. Learn. Res. , vol. 17, pp. 1–43, 2016. 7

2016
[14]

A variation al perspective on accelerated methods in optimization,

A. Wibisono, A. C. Wilson, and M. I. Jordan, “A variation al perspective on accelerated methods in optimization,” Proc. Natl. Acad. Sci. , vol. 113, no. 47, pp. E7351–E7358, 2016

2016
[15]

ADMM and accelerated ADMM as continuous dynamical systems,

G. Fran¸ ca, D. Robinson, and R. Vidal, “ADMM and accelerated ADMM as continuous dynamical systems,” 2018, arXiv:1805.06579

2018 arXiv
[16]

Understanding the acceleration phenomenon via high-resolution diﬀerent ial equations,

B. Shi, S. Du, M. I. Jordan, and W. Su, “Understanding the acceleration phenomenon via high-resolution diﬀerent ial equations,” 2018, arXiv:1810.08907

2018 arXiv
[17]

A dynamical systems perspective on Nesterov acceleration,

M. Muehlebach and M. I. Jordan, “A dynamical systems perspective on Nesterov acceleration,” 2019, arXiv:1905.07436

2019 arXiv
[18]

Inducing uniform asymptotic stability in time-varying accelerated optimization dynam ics via hybrid regularization,

J. I. Poveda and N. Li, “Inducing uniform asymptotic stability in time-varying accelerated optimization dynam ics via hybrid regularization,” 2019, arXiv:1905.12110

2019 arXiv
[19]

Some eﬀective methods for unconstrained optimization based on the solution of systems of ordinary diﬀerential equations,

A. Brown and M. Bartholomew-Biggs, “Some eﬀective methods for unconstrained optimization based on the solution of systems of ordinary diﬀerential equations,” J. Optimiz. Theory App. , vol. 62, no. 2, pp. 211–224, 1989

1989
[20]

A dynamical systems approach to constrained minimization,

J. Schropp and I. Singer, “A dynamical systems approach to constrained minimization,” Numer. Func. Anal. Opt., vol. 21, no. 3-4, pp. 537–551, 2000

2000
[21]

Direct Runge-Kutta discretization achieves acceleratio n,

J. Zhang, A. Mokhtari, S. Sra, and A. Jadbabaie, “Direct Runge-Kutta discretization achieves acceleratio n,” in Advances in Neural Information Processing Systems , 2018, pp. 3900–3909

2018
[22]

Analysis and desi gn of optimization algorithms via integral quadratic constrain ts,

L. Lessard, B. Recht, and A. Packard, “Analysis and desi gn of optimization algorithms via integral quadratic constrain ts,” SIAM J. Optim. , vol. 26, no. 1, pp. 57–95, 2016

2016
[23]

A uniﬁed analysis of stochastic optimization methods using jump system theory and quadratic constraints,

B. Hu, P. Seiler, and A. Rantzer, “A uniﬁed analysis of stochastic optimization methods using jump system theory and quadratic constraints,” in Proceedings of the 2017 Conference on Learning Theory , 2017, pp. 1157–1189

2017
[24]

Dissipativity theory for Nestero v’s accelerated method,

B. Hu and L. Lessard, “Dissipativity theory for Nestero v’s accelerated method,” in Proceedings of the 34th International Conference on Machine Learning , 2017, pp. 1549–1557

2017
[25]

Analysis of optimization algorithms via integral quadrat ic constraints: Nonstrongly convex problems,

M. Fazlyab, A. Ribeiro, M. Morari, and V. M. Preciado, “Analysis of optimization algorithms via integral quadrat ic constraints: Nonstrongly convex problems,” SIAM J. Optim. , vol. 28, no. 3, pp. 2654–2689, 2018

2018
[26]

Distribute d proximal augmented Lagrangian method for nonsmooth composite optimization,

S. Hassan-Moghaddam and M. R. Jovanovi´ c, “Distribute d proximal augmented Lagrangian method for nonsmooth composite optimization,” in Proceedings of the 2018 American Control Conference , Milwaukee, WI, 2018, pp. 2047–2052

2018
[28]

An exponentially convergent primal-dual algorithm for nonsmooth composite minimization,

D. Ding, B. Hu, N. K. Dhingra, and M. R. Jovanovi´ c, “An exponentially convergent primal-dual algorithm for nonsmooth composite minimization,” in Proceedings of the 57th IEEE Conference on Decision and Control , Miami, FL, 2018, pp. 4927–4932

2018
[29]

A control-theoretic approach to analysis and parameter sele ction of Douglas-Rachford splitting,

J. Seidman, M. Fazlyab, V. Preciado, and G. Pappas, “A control-theoretic approach to analysis and parameter sele ction of Douglas-Rachford splitting,” 2019, arXiv:1903.11525

2019 arXiv
[30]

Variance ampliﬁcation of accelerated ﬁrst-order algorit hms for strongly convex quadratic optimization problems,

H. Mohammadi, M. Razaviyayn, and M. R. Jovanovi´ c, “Variance ampliﬁcation of accelerated ﬁrst-order algorit hms for strongly convex quadratic optimization problems,” in Proceedings of the 57th IEEE Conference on Decision and Control, Miami, FL, 2018, pp. 5753–5758

2018
[31]

Performance of noisy Nesterov’s accelerated method for strongly convex optimization problems,

H. Mohammadi, M. Razaviyayn, and M. R. Jovanovi´ c, “Performance of noisy Nesterov’s accelerated method for strongly convex optimization problems,” in Proceedings of the 2019 American Control Conference , Philadelphia, PA, 2019, pp. 3426–3431

2019
[32]

Robustness of accelerated ﬁrst-order algorithms for stro ngly convex optimization problems,

H. Mohammadi, M. Razaviyayn, and M. R. Jovanovi´ c, “Robustness of accelerated ﬁrst-order algorithms for stro ngly convex optimization problems,” IEEE Trans. Automat. Control, 2019, submitted; also arXiv:1905.11011

2019 arXiv
[33]

Robust and structure exploiting optimization algorithms: An inte gral quadratic constraint approach,

S. Michalowsky, C. Scherer, and C. Ebenbauer, “Robust and structure exploiting optimization algorithms: An inte gral quadratic constraint approach,” 2019, arXiv:1905.00279

2019
[34]

System analysis via integ ral quadratic constraints,

A. Megretski and A. Rantzer, “System analysis via integ ral quadratic constraints,” IEEE Trans. Autom. Control, vol. 42, no. 6, pp. 819–830, 1997

1997
[35]

Gradient methods for minimizing functionals,

B. T. Polyak, “Gradient methods for minimizing functionals,” Zhurnal Vychislitel’noi Matematiki i Matematicheskoi Fiziki , vol. 3, no. 4, pp. 643–653, 1963

1963
[36]

Forward- backward truncated Newton methods for convex composite optimization,

P. Patrinos, L. Stella, and A. Bemporad, “Forward- backward truncated Newton methods for convex composite optimization,” 2014, arXiv:1402.6655

2014 arXiv
[37]

Forward–backward quasi-Newton methods for nonsmooth optimization problems,

L. Stella, A. Themelis, and P. Patrinos, “Forward–backward quasi-Newton methods for nonsmooth optimization problems,” Comput. Optim. Appl. , vol. 67, no. 3, pp. 443–487, 2017

2017
[38]

Forward-back ward envelope for the sum of two nonconvex functions: Further properties and nonmonotone line-search algorithms,

A. Themelis, L. Stella, and P. Patrinos, “Forward-back ward envelope for the sum of two nonconvex functions: Further properties and nonmonotone line-search algorithms,” SIAM J. Optim. , vol. 28, no. 3, pp. 2274–2303, 2018

2018
[39]

A second order primal-dual method for nonsmooth convex composite optimization,

N. K. Dhingra, S. Z. Khong, and M. R. Jovanovi´ c, “A second order primal-dual method for nonsmooth convex composite optimization,” IEEE Trans. Automat. Control , 2017, submitted; also arXiv:1709.01610

2017 arXiv
[40]

Nesterov, Introductory lectures on convex optimization: A basic course, 2013, vol

Y. Nesterov, Introductory lectures on convex optimization: A basic course, 2013, vol. 87

2013
[41]

Linear convergen ce of gradient and proximal-gradient methods under the Polyak - Lojasiewicz condition,

H. Karimi, J. Nutini, and M. Schmidt, “Linear convergen ce of gradient and proximal-gradient methods under the Polyak - Lojasiewicz condition,” in Joint European Conference on Machine Learning and Knowledge Discovery in Databases , 2016, pp. 795–811

2016
[42]

Exponential decay rate conditions for uncertain linear systems using integral quadratic constraints,

B. Hu and P. Seiler, “Exponential decay rate conditions for uncertain linear systems using integral quadratic constraints,” IEEE Trans. Autom. Control , vol. 61, no. 11, pp. 3631–3637, 2016

2016
[43]

On the numerical solution o f heat conduction problems in two and three space variables,

J. Douglas and H. Rachford, “On the numerical solution o f heat conduction problems in two and three space variables,” Trans. Amer. Math. Soc. , vol. 82, no. 2, pp. 421–439, 1956

1956
[44]

Linear convergence and metri c selection for Douglas-Rachford splitting and ADMM,

P. Giselsson and S. Boyd, “Linear convergence and metri c selection for Douglas-Rachford splitting and ADMM,” IEEE Trans. Automat. Control , vol. 62, no. 2, pp. 532–544, 2017

2017
[45]

On the Douglas-Rachfo rd splitting method and the proximal point algorithm for maximal monotone operators,

J. Eckstein and D. P. Bertsekas, “On the Douglas-Rachfo rd splitting method and the proximal point algorithm for maximal monotone operators,” Math. Program., vol. 55, no. 1-3, pp. 293–318, 1992

1992
[46]

Applications of the method of multipliers to variational inequalities,

D. Gabay, “Applications of the method of multipliers to variational inequalities,” in Studies in Mathematics and its Applications, 1983, vol. 15, pp. 299–331. 8

1983

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.

[1] [27]

On the exponential convergence rate of proximal gradient ﬂow algorithms,

S. Hassan-Moghaddam and M. R. Jovanovi´ c, “On the exponential convergence rate of proximal gradient ﬂow algorithms,” in Proceedings of the 57th IEEE Conference on Decision and Control , Miami, FL, 2018, pp. 4246–4251

work page 2018

[2] [1]

Distributed subgradient met hods for multiagent optimization,

A. Nedi´ c and A. Ozdaglar, “Distributed subgradient met hods for multiagent optimization,” IEEE Trans. on Automat. Control, vol. 54, no. 1, pp. 48–61, 2009

work page 2009

[3] [2]

A control perspective for centraliz ed and distributed convex optimization,

J. W ang and N. Elia, “A control perspective for centraliz ed and distributed convex optimization,” in Proceedings of the 50th IEEE Conference on Decision and Control , 2011, pp. 3800–3805

work page 2011

[4] [3]

New primal- dual proximal algorithm for distributed optimization,

P. Latafat, L. Stella, and P. Patrinos, “New primal- dual proximal algorithm for distributed optimization,” in Proceedings of the 55th IEEE Conference on Decision and Control, 2016, pp. 1959–1964

work page 2016

[5] [4]

A new randomized block-coordinate primal-dual proximal algori thm for distributed optimization,

P. Latafat, N. Freris, and P. Patrinos, “A new randomized block-coordinate primal-dual proximal algori thm for distributed optimization,” IEEE Trans. Automat. Control, 2019, doi:10.1109/TAC.2019.2906924

arXiv 2019

[6] [5]

A fast iterative shrinkage- thresholding algorithm for linear inverse problems,

A. Beck and M. Teboulle, “A fast iterative shrinkage- thresholding algorithm for linear inverse problems,” SIAM J. Imaging Sci. , vol. 2, no. 1, pp. 183–202, 2009

work page 2009

[7] [6]

Proximal algorithms,

N. Parikh and S. Boyd, “Proximal algorithms,” Found. Trends Opt., vol. 1, no. 3, pp. 123–231, 2013

work page 2013

[8] [7]

Studies in linear and non-linear programming,

K. J. Arrow, L. Hurwicz, and H. Uzawa, “Studies in linear and non-linear programming,” 1958

work page 1958