pith. sign in

arxiv: 2501.15419 · v3 · submitted 2025-01-26 · 🧮 math.OC

A primal-dual interior point trust region method for second-order stationary points of Riemannian inequality-constrained optimization problems

Mitsuaki Obara , Takayuki Okuno , Akiko Takeda This is my paper

Pith reviewed 2026-05-23 04:44 UTC · model grok-4.3

classification 🧮 math.OC
keywords Riemannian optimizationinterior-point methodstrust-region methodsinequality constraintssecond-order stationarityprimal-dual algorithmsglobal convergence
0
0 comments X

The pith

A primal-dual interior point trust region method converges globally to an approximate KKT point and a weak second-order stationary point for Riemannian inequality-constrained problems.

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

The paper develops a Riemannian primal-dual interior point trust region method for optimization problems that combine inequality constraints with a Riemannian manifold structure. It establishes that the iterates reach an approximate Karush-Kuhn-Tucker point together with a weak second-order stationary point, and that the result strengthens to full second-order stationarity when strict complementarity holds. The construction merges an interior-point barrier with trust-region subproblems posed in the tangent space so that each step respects the manifold geometry while driving the barrier parameter to zero. This combination is relevant for nonconvex problems in control and machine learning because first-order conditions alone often fail to identify local minima on curved domains.

Core claim

The Riemannian primal-dual interior point trust region method (RIPTRM) is proved to converge globally to an approximate Karush-Kuhn-Tucker point and a weak second-order stationary point. Under the strict complementarity condition the convergence strengthens to a second-order stationary point. The algorithm is the first to combine a trust-region strategy with an interior-point framework for constrained optimization on Riemannian manifolds while retaining the second-order stationarity guarantee for nonlinear inequality constraints.

What carries the argument

The trust-region subproblem solved at each iteration, which incorporates the primal-dual interior-point barrier conditions into a quadratic model on the tangent space of the manifold.

If this is right

  • The method applies directly to nonconvex problems with inequality constraints that arise in control and machine learning on manifolds.
  • Both a truncated conjugate gradient solver and an eigenvalue-based exact solver can be used for the trust-region subproblems.
  • Numerical tests show that the algorithm consistently reaches solutions of high accuracy.
  • The exact subsolver variant performs well on instances where the Hessian of the Lagrangian has a large negative eigenvalue.

Where Pith is reading between the lines

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

  • The same subproblem structure could be reused to design line-search variants that still guarantee second-order stationarity.
  • Replacing the fixed barrier with an adaptive one might reduce the number of outer iterations without losing the convergence guarantees.
  • The eigenvalue-based subsolver already identifies negative curvature; extending it to a negative-curvature escape step would connect the method to saddle-point avoidance techniques.

Load-bearing premise

The analysis requires the manifold and the objective and constraint functions to be sufficiently smooth and to satisfy standard constraint qualifications that keep the interior-point and trust-region steps well-defined.

What would settle it

A concrete sequence of iterates generated by the method that converges to a point violating the second-order necessary conditions, while all stated smoothness and qualification assumptions continue to hold, would falsify the global convergence claim.

read the original abstract

We consider Riemannian inequality-constrained optimization problems. Such problems inherit the benefits of Riemannian approach developed in the unconstrained setting and naturally arise from applications in control, machine learning, and other fields. We propose a Riemannian primal-dual interior point trust region method (RIPTRM) for solving them. We prove its global convergence to an approximate Karush-Kuhn-Tucker point and a weak second-order stationary point. Under the strict complementarity condition, this result reduces to global convergence to a second-order stationary point. To the best of our knowledge, this is the first algorithm that incorporates the trust region strategy for constrained optimization on Riemannian manifolds, and has the second-order convergence property for optimization problems on Riemannian manifolds with nonlinear inequality constraints. We conduct numerical experiments in which we introduce a truncated conjugate gradient method and an eigenvalue-based subsolver for RIPTRM to approximately and exactly solve the trust region subproblems, respectively. Empirical results show that RIPTRMs consistently find solutions with high accuracy. Additionally, we observe that RIPTRM with the exact search direction shows promising performance in an instance where the Hessian of the Lagrangian has a large negative eigenvalue.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

0 major / 3 minor

Summary. The paper proposes the Riemannian primal-dual interior point trust region method (RIPTRM) for inequality-constrained optimization on Riemannian manifolds. It establishes global convergence of the method to an approximate KKT point and a weak second-order stationary point; under strict complementarity this strengthens to convergence to a second-order stationary point. The work also describes numerical experiments that employ a truncated conjugate gradient method and an eigenvalue-based subsolver to solve the trust-region subproblems approximately and exactly, respectively, with results indicating high-accuracy solutions.

Significance. If the convergence analysis is rigorous, the result is significant: it supplies the first trust-region algorithm for Riemannian problems with nonlinear inequality constraints that achieves second-order stationarity guarantees. The explicit treatment of both approximate and exact subproblem solvers, together with the reduction from weak to full second-order stationarity under strict complementarity, constitutes a clear technical contribution. The standard smoothness and constraint-qualification assumptions are stated explicitly, and the manuscript supplies the full convergence argument rather than an outline.

minor comments (3)
  1. [Numerical Experiments] Numerical experiments section: the reported results are described at a high level without error bars, multiple random seeds, or quantitative baseline comparisons against existing Riemannian interior-point or penalty methods; this weakens the empirical support for the practical claims.
  2. [Introduction] The statement that RIPTRM is the first such method should be accompanied by a concise table or paragraph in the introduction that directly contrasts it with the closest prior Riemannian constrained solvers (e.g., those using penalty or augmented Lagrangian approaches) to make the novelty claim precise.
  3. [Preliminaries] Notation for the Riemannian metric, retraction, and vector transport is introduced without a dedicated preliminary subsection; a short table collecting these symbols and their properties would improve readability for readers outside the immediate subfield.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for their positive summary of the manuscript, recognition of its technical contributions, and recommendation of minor revision. We are pleased that the work is viewed as supplying the first trust-region primal-dual interior-point method on Riemannian manifolds with second-order stationarity guarantees.

Circularity Check

0 steps flagged

No significant circularity identified

full rationale

The paper presents a convergence proof for a new Riemannian primal-dual interior-point trust-region algorithm. The central claims rest on standard smoothness, constraint qualification, and strict complementarity assumptions that are stated explicitly and are independent of the algorithm's outputs. No derivation step reduces a claimed prediction or stationarity result to a fitted parameter, self-referential definition, or load-bearing self-citation chain; the proof structure follows conventional non-circular analysis for trust-region and interior-point methods on manifolds. The 'first algorithm' claim is a novelty statement, not a mathematical reduction.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The central claims rest on standard Riemannian manifold assumptions and constraint qualifications that are not derived in the paper.

axioms (2)
  • domain assumption The feasible set is a Riemannian manifold with the objective and constraint functions twice continuously differentiable.
    Invoked to define the KKT conditions and trust-region subproblems.
  • domain assumption Standard constraint qualifications (e.g., LICQ or MFCQ) hold at limit points.
    Required for the KKT stationarity statements.

pith-pipeline@v0.9.0 · 5740 in / 1295 out tokens · 23782 ms · 2026-05-23T04:44:51.126669+00:00 · methodology

discussion (0)

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

Forward citations

Cited by 1 Pith paper

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

  1. Local near-quadratic convergence of Riemannian interior point methods

    math.OC 2025-05 unverdicted novelty 7.0

    Riemannian interior point methods achieve local near-quadratic convergence to second-order stationary points, with the first algorithm shown to combine this local rate and global convergence guarantees.

Reference graph

Works this paper leans on

71 extracted references · 71 canonical work pages · cited by 1 Pith paper · 2 internal anchors

  1. [1]

    Absil, R

    P.-A. Absil, R. Mahony, and R. Sepulchre , Optimization algorithms on matrix manifolds , Princeton University Press, Princeton, 2008

    work page 2008

  2. [2]

    Adachi, S

    S. Adachi, S. Iwata, Y. Nakatsukasa, and A. Takeda , Solving the trust-region subproblem by a generalized eigenvalue problem , SIAM J. Optim., 27 (2017), pp. 269–291

    work page 2017

  3. [3]

    Agarwal, N

    N. Agarwal, N. Boumal, B. Bullins, and C. Cartis , Adaptive regularization with cubics on manifolds, Math. Program., 188 (2021), pp. 85–134

    work page 2021

  4. [4]

    Andreani, K

    R. Andreani, K. R. Couto, O. P. Ferreira, and G. Haeser , Constraint qualifications and strong global convergence properties of an augmented Lagrangian method on Riemannian manifolds, SIAM J. Optim., 34 (2024), pp. 1799–1825

    work page 2024

  5. [5]

    Andreani, K

    R. Andreani, K. R. Couto, O. P. Ferreira, G. Haeser, and L. F. Prudente , Global con- vergence of an augmented Lagrangian method for nonlinear programming via Riemannian optimization, 2024, https://optimization-online.org/?p=27595

    work page 2024

  6. [6]

    Arahata, T

    S. Arahata, T. Okuno, and A. Takeda , Complexity analysis of interior-point methods for second-order stationary points of nonlinear semidefinite optimization problems , Comput. Optim. Appl., 86 (2023), pp. 555–598

    work page 2023

  7. [7]

    H. Y. Benson, R. J. V anderbei, and D. F. Shanno , Interior-point methods for nonconvex nonlinear programming: filter methods and merit functions , Comput. Optim. Appl., 23 (2002), pp. 257–272

    work page 2002

  8. [8]

    Bergmann and R

    R. Bergmann and R. Herzog , Intrinsic formulation of KKT conditions and constraint qual- ifications on smooth manifolds , SIAM J. Optim., 29 (2010), pp. 2423–2444

    work page 2010

  9. [9]

    E. G. Birgin, O. P. Ferreira, G. Haeser, N. Maculan, L. M. Ramirez, and L. F. Pru- dente, Smoothing ℓ1-exact penalty methiod for intrinsically constarined Riemannian op- timization problems, 2025, https://optimization-online.org/?p=28986

    work page 2025

  10. [10]

    Boumal, An introduction to optimization on smooth manifolds, Cambridge University Press, Cambridge, 2023

    N. Boumal, An introduction to optimization on smooth manifolds, Cambridge University Press, Cambridge, 2023

    work page 2023

  11. [11]

    Boumal, P.-A

    N. Boumal, P.-A. Absil, and C. Cartis , Global rates of convergence for nonconvex optimiza- tion on manifolds , IMA J. Numer. Anal., 39 (2019), pp. 1–33

    work page 2019

  12. [12]

    Brossette, A

    S. Brossette, A. Escande, and A. Kheddar , Multicontact postures computation on mani- folds, IEEE Trans. Robotics, 34 (2018), pp. 1252–1265

    work page 2018

  13. [13]

    R. H. Byrd, J. C. Gilbert, and J. Nocedal , A trust region method based on interior point techniques for nonlinear programming, Math. Program., 89 (2000), pp. 149–185

    work page 2000

  14. [14]

    R. H. Byrd, G. Liu, and J. Nocedal , On the local behavior of an interior point method for nonlinear programming, in Numerical Analysis 1997, Chapman and Hall/CRC, Harlow, 1997, pp. 37–56

    work page 1997

  15. [15]

    Canary, D

    R. Canary, D. Epstein, and A. Marden , Fundamentals of hyperbolic manifolds: selected expositions, Cambridge University Press, Cambridge, 2006

    work page 2006

  16. [16]

    Cartis, N

    C. Cartis, N. I. M. Gould, and P. L. Toin , Evaluation complexity of algorithms for non- convex optimization, SIAM, Philadelphia, 2022

    work page 2022

  17. [17]

    A. R. Conn, N. I. M. Gould, D. Orban, and P. L. Toint , A primal-dual trust-region algorithm for non-convex nonlinear programming , Math. Program., 87 (2000), pp. 215– 249

    work page 2000

  18. [18]

    A. R. Conn, N. I. M. Gould, and P. L. Toint , Trust region methods, SIAM, Philadelphia, 2000

    work page 2000

  19. [19]

    Criscitiello and N

    C. Criscitiello and N. Boumal , Efficiently escaping saddle points on manifolds , in NeurIPS, 2019, pp. 5987–5997

    work page 2019

  20. [20]

    F. E. Curtis, N. I. M. Gould, D. P. Robinson, and P. L. Toint , An interior-point trust- funnel algorithm for nonlinear optimization , Math. Program., 161 (2017), pp. 73–134

    work page 2017

  21. [21]

    F. R. de Oliveira and O. P. Ferreira , Newton method for finding a singularity of a special class of locally Lipschitz continuous vector fields on Riemannian manifolds , J. Optim. Theory Appl., 185 (2020), pp. 522–539

    work page 2020

  22. [22]

    M. P. do Carmo , Riemannian Geometry, Birkh¨ auser, Basel, 1992

    work page 1992

  23. [23]

    Forsgren, P

    A. Forsgren, P. E. Gill, and M. H. Wright , Interior methods for nonlinear optimization , SIAM Rev., 44 (2002), pp. 525–597

    work page 2002

  24. [24]

    N. I. M. Gould, D. Orban, A. Sartenaer, and P. L. Toint , Superlinear convergence of primal-dual interior point algorithms for nonlinear programming , SIAM J. Optim., 11 (2001), pp. 974–1002

    work page 2001

  25. [25]

    N. I. M. Gould, D. Orban, and P. L. Toint , Numerical methods for large-scale nonlinear 26 M. OBARA, T. OKUNO, AND A. TAKEDA optimization, Acta Numer., 14 (2005), pp. 299–361

    work page 2005

  26. [26]

    Goyens and C

    F. Goyens and C. W. Royer , Riemannian trust-region methods for strict saddle functions with complexity guarantees. arXiv:2402.07614v2, 2024

    work page arXiv 2024

  27. [27]

    Hinder and Y

    O. Hinder and Y. Ye , Worst-case iteration bounds for log barrier methods on problems with nonconvex constraints, Math. Oper. Res., (2023). to appear

    work page 2023

  28. [28]

    Hirai, H

    H. Hirai, H. Nieuwboer, and M. W alter , Interior-point methods on manifolds: theory and applications, in FOCS, 2023, pp. 2021–2030

    work page 2023

  29. [29]

    Huang, P.-A

    W. Huang, P.-A. Absil, and K. A. Gallivan , A Riemannian symmetric rank-one trust-region method, Math. Program., 150 (2015), pp. 179–216

    work page 2015

  30. [30]

    Huang, P.-A

    W. Huang, P.-A. Absil, and K. A. Gallivan , A Riemannian BFGS method without dif- ferentiated retraction for nonconvex optimization problems , SIAM J. Optim., 28 (2018), pp. 470–495

    work page 2018

  31. [31]

    Kasai and B

    H. Kasai and B. Mishra , Inexact trust-region algorithms on Riemannian manifolds , in NeurIPS, 2018, pp. 4252–4265

    work page 2018

  32. [32]

    Kiessling, S

    D. Kiessling, S. Leyffer, and C. V anaret , A unified funnel restoration SQP algorithm . arXiv:2404.09208, 2024

    work page arXiv 2024

  33. [33]

    Lai and A

    Z. Lai and A. Yoshise , Riemannian interior point methods for constrained optimization on manifolds, J. Optim. Theory Appl., 201 (2024), pp. 433–469

    work page 2024

  34. [34]

    J. M. Lee , Introduction to smooth manifolds , Springer, New York, second ed., 2012

    work page 2012

  35. [35]

    J. M. Lee , Introduction to Riemannian manifolds , Springer, New York, second ed., 2018

    work page 2018

  36. [36]

    Levin, J

    E. Levin, J. Kileel, and N. Boumal , The effect of smooth parameterizations on nonconvex optimization landscapes, Math. Program., (2024). to appear

    work page 2024

  37. [37]

    Liu and N

    C. Liu and N. Boumal , Simple algorithms for optimization on Riemannian manifolds with constraints, Appl. Math. Optim., 82 (2020), pp. 949–981

    work page 2020

  38. [38]

    Y. Luo, X. Li, and A. Zhang , Nonconvex factorization and manifold formulations are almost equivalent in low-rank matrix optimization , INFORMS J. Optim., (2024). to appear

    work page 2024

  39. [39]

    Luo and N

    Y. Luo and N. G. Trillos , Nonconvex matrix factorization is geodesically convex: global landscape analysis for fixed-rank matrix optimization from a Riemannian perspective . arXiv:2209.15130, 2022

    work page arXiv 2022

  40. [40]

    Narushima, S

    Y. Narushima, S. Nakayama, M. Takemura, and H. Yabe , Memoryless quasi-Newton meth- ods based on the spectral-scaling Broyden family for Riemannian optimization , J. Optim. Theory Appl., 197 (2023), pp. 639–664

    work page 2023

  41. [41]

    Nocedal and S

    J. Nocedal and S. J. Wright , Numerical Optimization, Springer, New York, 2006

    work page 2006

  42. [42]

    Obara, T

    M. Obara, T. Okuno, and A. Takeda , Sequential quadratic optimization for nonlinear opti- mization problems on Riemannian manifolds , SIAM J. Optim., 32 (2022), pp. 822–853

    work page 2022

  43. [43]

    Local near-quadratic convergence of Riemannian interior point methods

    M. Obara, T. Okuno, and A. Takeda , Local near-quadratic convergence of Riemannian interior point methods. arXiv:2505.19724, 2025

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  44. [44]

    Obara, K

    M. Obara, K. Sato, H. Sakamoto, T. Okuno, and A. Takeda , Stable linear system iden- tification with prior knowledge by Riemannian sequential quadratic optimization , IEEE Trans. Automat. Control, 69 (2024), pp. 2060–2066

    work page 2024

  45. [45]

    Sakai and H

    H. Sakai and H. Iiduka , Modified memoryless spectral-scaling Broyden family on Riemannian optimization, J. Optim. Theory Appl., 202 (2024), pp. 834–853

    work page 2024

  46. [46]

    Sato, Riemannian conjugate gradient methods: general framework and specific algorithms with convergence analyses, SIAM J

    H. Sato, Riemannian conjugate gradient methods: general framework and specific algorithms with convergence analyses, SIAM J. Optim., 32 (2022), pp. 2690–2717

    work page 2022

  47. [47]

    Sato and K

    H. Sato and K. Sato , Riemannian optimal system identification algorithm for linear MIMO systems, IEEE Control Syst. Lett., 1 (2017), pp. 376 – 381

    work page 2017

  48. [48]

    Sato and H

    K. Sato and H. Sato , Structure-preserving H 2 optimal model reduction based on the Rie- mannian trust-region method, IEEE Trans. Automat. Control, 63 (2017), pp. 505–512

    work page 2017

  49. [49]

    K. Sato, H. Sato, and T. Damm , Riemannian optimal identification method for linear sys- tems with symmetric positive-definite matrix , IEEE Trans. Automat. Control, 65 (2020), pp. 4493–4508

    work page 2020

  50. [50]

    Schiela and J

    A. Schiela and J. Ortiz , An SQP method for equality constrained optimization on Hilbert manifolds, SIAM J. Optim., 31 (2021), pp. 949–981

    work page 2021

  51. [51]

    Silva, M

    R. Silva, M. Ulbrich, S. Ulbrich, and L. N. Vicente , A globally convergent primal-dual interior-point filter method for nonlinear programming: new filter optimality measures and computational results, 2008, http://hdl.handle.net/10316/11218

    work page 2008

  52. [52]

    Sra and R

    S. Sra and R. Hosseini , Conic geometric optimization on the manifold of positive definite matrices, SIAM J. Optim., 25 (2015), pp. 713–739

    work page 2015

  53. [53]

    S. Sra, N. K. Vishnoi, and O. Yildiz , On geodesically convex formulations for the Brascamp- Lieb constant, in APPROX/RANDOM, vol. 116, 2018, pp. 25:1–25:15

    work page 2018

  54. [54]

    Y. Sun, N. Flammarion, and M. F azel , Escaping from saddle points on Riemannian mani- folds, in NeurIPS, 2019, pp. 7276–7286. RIEMANNIAN INTERIOR POINT TRUST REGION METHOD 27

    work page 2019

  55. [55]

    Townsend, N

    J. Townsend, N. Koep, and S. Weichwald , Pymanopt: A python toolbox for optimization on manifolds using automatic differentiation , J. Mach. Learn. Res., 17 (2016), pp. 1—-5

    work page 2016

  56. [56]

    Tseng, Convergent infeasible interior-point trust-region methods for constrained minimiza- tion, SIAM J

    P. Tseng, Convergent infeasible interior-point trust-region methods for constrained minimiza- tion, SIAM J. Optim., 13 (2002), pp. 432–469

    work page 2002

  57. [57]

    Ulbrich, S

    M. Ulbrich, S. Ulbrich, and L. N. Vicente , A globally convergent primal-dual interior-point filter method for nonlinear programming, Math. Program., 100 (2004), pp. 379–410

    work page 2004

  58. [58]

    W ¨achter and L

    A. W ¨achter and L. T. Biegler , Failure of global convergence for a class of interior point methods for nonlinear programming, Math. Program., 88 (2000), pp. 565–574

    work page 2000

  59. [59]

    Weber and S

    M. Weber and S. Sra , Global optimality for Euclidean CCCP under Riemannian convexity , in ICML, 2023, pp. 202:36709–36803

    work page 2023

  60. [60]

    Yamakawa and H

    Y. Yamakawa and H. Sato , Sequential optimality conditions for nonlinear optimization on Riemannian manifolds and a globally convergent augmented Lagrangian method , Comput. Optim. Appl., 81 (2022), pp. 397–421

    work page 2022

  61. [61]

    Yamashita, H

    H. Yamashita, H. Yabe, and T. Tanabe , A globally and superlinearly convergent primal- dual interior point trust region method for large scale constrained optimization , Math. Program., 102 (2005), pp. 111–151

    work page 2005

  62. [62]

    W. H. Yang, L.-H. Zhang, and R. Song , Optimality conditions for the nonlinear programming problems on Riemannian manifolds , Pac. J. Optim., 10 (2014), pp. 415–434

    work page 2014

  63. [63]

    K. Ye, K. S.-W. Wong, and L.-H. Lim , Optimization on flag manifolds , Math. Program., 194 (2022), pp. 621–660

    work page 2022

  64. [64]

    Zass and A

    R. Zass and A. Shashua , Nonnegative sparse PCA, in NIPS, 2006, pp. 1561–1568

    work page 2006

  65. [65]

    J. Z. S. Zhang , A cubic regularized Newton’s method over Riemannian manifolds . arXiv:1805.05565, 2018. Appendix A. Proofs of Lemmas 4.8 and 4.13. In this section, we derive the proofs of the key lemmas, Lemmas 4.8 and 4.13, that estimate the gaps between the predicted and actual reductions. To prove the lemmas, we first express the gaps in the following...

    work page internal anchor Pith review Pith/arXiv arXiv 2018

  66. [66]

    Then, Assumption B.7 holds

    Suppose R = Exp and that, for every i ∈ I , the function gi : M → R is Li-Lipschitz continuous; that is, there exists Li > 0 such that |gi(x1) − gi(x2)| ≤ Li dist(x1, x2) for all x1, x2 ∈ M. Then, Assumption B.7 holds

    work page

  67. [67]

    If xℓ ℓ is bounded, then Assumption B.7 holds. Proof. We first prove Item 1. Notice that, due to the completeness of M, the domain of the exponential map atx is the whole of Tx M for all x ∈ M. It follows from [10, Proposition 10.41] that Li-Lipschitz continuity is equivalent to gi Expx(ξx) − gi(x) ≤ Li∥ξx ∥x for all x ∈ M and all ξx ∈ Tx M. Therefore, fo...

    work page

  68. [68]

    If xℓ ℓ is bounded, then Assumption B.10 holds

    work page

  69. [69]

    If M is compact, then Assumption B.10 holds. Proof. Under Item 1 or Item 2, let Q ⊆ M be a compact subset with xℓ ℓ ⊆ Q. For any δR > 0, define T := (x, ξx) ∈ TM: x ∈ Q and ∥ξx ∥x ≤ δR . Note that T is compact [10, Exercise 10.31]. Thus, from the smoothness of R and the continuity of the operator norm, there exists LR > 0 such that ∥D Rx(ξx)∥op ≤ LR for a...

    work page

  70. [70]

    Assumption B.8 holds

    work page

  71. [71]

    Under Assumptions B.5 and B.7, if lim j→∞ dx ℓj x ℓj = 0,(E-3) then Assumption B.9 holds. Proof. First, we consider Item 1. Since the dual variable is updated when the primal iterate is successful, we focus on the successful iterations. From Item 1 of Lemma 4.11, there exists ˜ε > 0 such that 1 ˜ ε ≥ 1 gi(x ℓ) for any i ∈ I and all ℓ ∈ N0, which, together...

    work page