Pith. sign in

REVIEW 3 major objections 5 minor 66 references

A globally and superlinearly convergent QO-free method for nonlinear optimization on Riemannian manifolds

T0 review · 3 major / 5 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read This paper proposes the first quadratic-optimization-free method for constrained optimization on Riemannian manifolds and proves global, strong, and Q-superlinear convergence under standard assumptions.

desk verdict Solid global convergence for a first Riemannian QO-free method, but the superlinear/Maratos claim rests on a second-order retraction assumption that the paper's own experiments violate. read the letter →

arxiv 2608.09755 v1 pith:36RARAYA submitted 2026-08-10 math.OC

Chunming Tang , Hao He , Wen Huang , Jinbao Jian , Ruyun Li This is my paper · ORCID
classification math.OC MSC 90C3090C48
keywords ConstrainedoptimizationRiemannianmanifoldsQO-freemethodGlobalconvergenceSuperlinearMaratoseffectsecond-orderretraction
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

Given a smooth cost and smooth equality and inequality constraints on a complete Riemannian manifold, the paper wants a method that always solves feasible subproblems, is globally convergent, and converges Q-superlinearly near a solution. It extends the Euclidean quadratic-optimization-free (QO-free) approach, which solves linear systems instead of quadratic programs, to the manifold setting. The algorithm starts from a strictly feasible point, builds a master tangent direction from three linear systems with one shared operator, adds a higher-order correction direction, and performs an arc search. The paper proves that the whole sequence converges to a KKT point and that, near a solution satisfying strict complementarity and second-order sufficient conditions, the unit step is eventually accepted, giving superlinear convergence. If correct, this is the first constrained manifold method that provably overcomes the Maratos effect rather than assuming it away.

What carries the argument

The carrier of the argument is the linear operator $A_k: T_{x_k}M\times\mathbb{R}^{m+\ell}\to T_{x_k}M\times\mathbb{R}^{m+\ell}$ defined by $$A_k(\eta,\$\lambda$)=\left(H_k[\eta]+\sum_{i\in L}\lambda_i\,\mathrm{grad}\,c_i(x_k),\ \alpha_i^k\langle\mathrm{grad}\,c_i(x_k),\eta\rangle-\sqrt{2}\beta_i^k\lambda_i\right),$$ whose nonsingularity makes the three search-direction systems (10), (14), and (15) uniquely solvable with one shared factorization. The coefficients $\alpha_i^k$ and $\beta_i^k$ come from the complementarity function $\phi(a,b)=\sqrt{a^2+b^2}-a-b$ and encode the complementarity conditions. Around this, the method uses the semi-penalty reformulation $F_\rho=f-\rho\sum_{i\in E}c_i$, a $\theta_k$ combination rule that balances descent and feasibility, and a higher-order correction direction from the linear least-squares subproblem (18). The superlinear proof transfers the iteration to Riemannian normal coordinates and uses a second-order retraction to control the second-order terms in the arc-search expansion.

What would settle it

Run Algorithm 1 on a small constrained problem on the sphere or the oblique manifold using the exponential map as the retraction, initialize near a KKT point satisfying strict complementarity and SOSC, and record the sequence of ratios $\|x_{k+1}-x_*\|/\|x_k-x_*\|$. The theorem predicts these ratios tend to zero; observing a single run where they are bounded away from zero while the stated assumptions hold would disprove the superlinear claim.

Watch

Extended reading notes

Core claim

The central discovery is that the QO-free mechanism is compatible with Riemannian geometry and with equality constraints, provided the equality constraints are treated by a semi-penalty reformulation and the retraction is sufficiently accurate. Algorithm 1 solves three linear systems that share the same nonsingular operator $A_k$, combines their solutions to obtain a feasible descent direction $\eta_k$, and then solves a reduced least-squares problem for a correction $\tilde{\eta}_k$. The paper proves under Assumptions 4.1--4.4 that every accumulation point of the iterates is a KKT point of the original problem, and under Assumptions 5.1--5.5 that the entire sequence converges and satisfies $\|\hat{x}_{k+1} - \hat{x}_*\| = o(\|\hat{x}_k - \hat{x}_*\|)$.

Load-bearing premise

The load-bearing premise for the fast-convergence claim is Assumption 5.4: the retraction must be second-order, matching the exponential map up to second order along every tangent direction, because the proof's Taylor expansion near the solution has no other way to control the second-order term. If a practical retraction satisfies only the usual first-order property, the unit-step acceptance and hence the superlinear rate are not guaranteed by this paper.

Editorial extensions

If this is right

  • Every iterate is strictly feasible for the penalized problem $P_\rho$, so the method never hits an infeasible subproblem; this removes a failure mode present in earlier Riemannian SQP methods.
  • Solving three linear systems with a common operator costs essentially one factorization, making each outer iteration inexpensive when the manifold dimension is moderate.
  • The penalty parameter $\rho_k$ is increased only finitely many times; afterwards the method behaves as a feasible descent method and its accumulation points are KKT points of the original problem.
  • Near a solution satisfying strict complementarity, SOSC, the Hessian approximation condition, and a second-order retraction, the step size $t_k=1$ is eventually accepted, so the fast local convergence is not blocked by the Maratos effect.

Reading between the lines

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

  • A natural stress test is to track the ratio $\|x_{k+1}-x_*\|/\|x_k-x_*\|$ on a matrix manifold equipped with a first-order projection retraction; if the unit step is still accepted and the ratio tends to zero, then Assumption 5.4 is not numerically essential for the local result.
  • The same proof architecture should transfer to active-set variants, since the correction subproblem is activated only near the solution; that would reduce per-iteration cost on problems with many constraints.
  • If the strict-feasibility initialization is replaced by a strongly subfeasible direction phase, the method could be extended to problems whose feasible sets are hard to enter from outside.
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.

Referee Report

3 major / 5 minor

Summary. The paper extends Euclidean QO-free (quadratic-optimization-free) methods to nonlinearly constrained optimization on Riemannian manifolds. Problem (P) is converted via a semi-penalty scheme into an inequality-constrained problem (P_ρ), and Algorithm 1 generates strictly feasible iterates by solving three linear systems with a common operator, a reduced least-squares correction subproblem, and a Riemannian arc search. The authors prove global convergence (Theorem 4.1), strong convergence (Theorem 5.1), eventual acceptance of the unit step size (Theorem 5.2), and Q-superlinear convergence (Theorem 5.3) under assumptions including LICQ, SOSC, a Dennis–Moré condition on the Hessian approximations, and a second-order retraction assumption. Numerical experiments on nonnegative low-rank matrix completion and nonnegative PCA compare the method with RALM, REPMs, and RSQO, reporting competitive accuracy and speed.

Significance. If the convergence claims hold, this is a meaningful contribution: it provides a feasible QO-free method on manifolds that avoids solving quadratic subproblems at each iteration, achieves global and strong convergence, and addresses the Maratos effect without imposing it as an assumption. The algorithmic device of three linear systems sharing one operator, plus a higher-order correction direction solved by a least-squares subproblem, is a natural and potentially useful extension of the Euclidean theory. The paper is largely rigorous and self-contained for the main theorems, although some proof details are deferred to the companion Euclidean results. The numerical experiments are reproducible in structure and show competitive performance. The main qualification is that the superlinear convergence and Maratos-effect statements are established only under Assumption 5.4, which is not satisfied by the retractions used in the reported experiments; this limits the scope of the headline claim and needs to be addressed.

major comments (3)
  1. [Section 5, Assumption 5.4 and Theorem 5.2; Section 6] The proof of Theorem 5.2 relies critically on the second-order retraction assumption: in the Taylor expansion (39), the term involving the second covariant derivative of the retraction is dropped, and the subsequent cancellation leading to (48) uses that this term is zero. For a general first-order retraction, this term is generically O(||η_k||^2) and can be positive, so the unit-step acceptance (19) is not guaranteed. The retractions used in the numerical experiments — the elementwise normalization retraction on the oblique manifold and the SVD-truncation projection on the fixed-rank manifold — are only first-order and generically have nonzero second derivative. Consequently, the superlinear convergence shown in Figure 1 and Tables 1–5 is not covered by Theorem 5.3. Please either prove the result for first-order retractions (e.g., by modifying the correction or arc-search mechanism), run experiments with second-order retractions (such as exponential maps on manifolds where they are tractable), or explicitly state in the abstract and conclusions that the Maratos-effect and superlinear convergence guarantees require second-order retractions and that the experiments are heuristic for that aspect.
  2. [Section 4, Lemma 4.2] The proof of Lemma 4.2 is omitted: the text says the boundedness of {||A_k^{-1}||_op} follows the same lines as Lemma 3.1 in [14]. This lemma is load-bearing for Lemma 4.3 and hence for the global convergence Theorem 4.1, and in the manifold setting one must contend with a sequence of operators acting on different tangent spaces. Please provide a self-contained proof or a precise statement of how the Euclidean proof is adapted to the Riemannian setting, including the role of Assumptions 4.1–4.4 in controlling the spectral norm of the inverse operators.
  3. [Section 5, Theorem 5.3, Eq. (56)] The key estimate ||η_{k0} + o(||η_{k0}||)||^2 = o(||x̂_k − x̂_*||) is asserted by analogy with [14, Lemma 4.8], without proof. This estimate is the bridge between the master direction and the local error and is essential for the Q-superlinear rate. The analogy is not immediate because of the presence of the normal-coordinate chart, the metric G_{x̂_k}, and the retraction R̂. Please supply the full derivation in this paper, or at least verify explicitly that each step in the Euclidean proof carries over to the current setting.
minor comments (5)
  1. [Section 3.1, Eq. (11)] In the sentence following Eq. (11), “It has been shown in [49, 50] that αk k and βk i satisfy” should read “α_k^i and β_k^i”; the current text contains a typo.
  2. [Section 5, Lemma 5.9 and Theorem 5.3] In Lemma 5.9 and Theorem 5.3, the notation “o‖x̂_k − x̂_∗‖” and “o(‖x̂_k − x̂_∗‖” is missing parentheses or has unbalanced delimiters; please write o(‖x̂_k − x̂_∗‖) consistently.
  3. [Section 6.3, Figure 1 caption and text] The text says “ROQ-free” in the description of Figure 1; this should be “RQO-free”.
  4. [Section 6.2] The method name is spelled inconsistently as “RSQO” in most places and “RSQP” in one sentence; please standardize the abbreviation.
  5. [Section 5, Theorem 5.2, Eq. (57)] Equation (57) and the following display contain an unmatched closing parenthesis in ‖R̂_{x̂}(η̂_{k0}+o(‖η̂_{k0}‖)) − (x̂_k + η̂_{k0}+o(‖η̂_{k0}‖)‖; the typography should be corrected.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation: global and superlinear convergence follow from stated hypotheses and external QO-free theory; the second-order retraction and Dennis-Moré assumptions are coverage gaps, not circular inputs.

full rationale

Walking the paper's derivation chain, I found no step in which a claimed prediction reduces by construction to fitted inputs or to a self-citation. The master direction is defined by the linear systems (10), (14), (15), and (16), and Lemmas 3.1-3.3 prove well-definedness and descent using the positive definiteness of H_k and the Fischer-Burmeister scalars (7); none of these lemmas imports the target convergence result. Global convergence (Theorem 4.1) is derived from Assumptions 4.1-4.4 with proofs following the external references [14] and [48], not from the authors' own prior results. Strong and superlinear convergence (Theorems 5.1-5.3) are conditional on strict complementarity, SOSC, second-order retraction (Assumption 5.4), and the Dennis-Moré condition (Assumption 5.5); these are stated hypotheses, not restatements of the conclusion, and the local reduction to [14] is an external comparison. The higher-order correction (18) and the definition of the scalar ̟_k are selected so that the unit step can be verified to satisfy (19)-(20); Theorem 5.2 actively checks those inequalities rather than assuming them. Author self-citations ([9], [10], [26], [59], [60]) appear only in introductory or future-work contexts and are not load-bearing premises of the convergence theorems. The numerical experiments use Manopt retractions that may be only first-order, so Assumption 5.4 may be violated in the tested implementation; however, that is a coverage gap, not a circular reduction. No fitted parameter is re-labeled as a prediction, and no equation is equivalent to its input by definition.

Assumptions & free parameters 4 free parameters · 7 assumptions · 0 invented entities

No new physical or mathematical entities are postulated. The method itself is a new algorithm, but it is composed of existing mathematical objects (linear operators, retractions, Fischer-Burmeister functions).

free parameters (4)
  • ν = 2.3
    Exponent in the feasibility-enhancing term; theory requires ν > 2, numerical value chosen by hand.
  • τ = 0.75
    Weight balancing descent and feasibility in θ_k; theory requires τ ∈ (0,1), numerical value chosen by hand.
  • σ = 0.45
    Sufficient decrease parameter in line search; theory requires σ ∈ (0,1/2), numerical value chosen by hand.
  • ρ0, ρ̃, r1, r2, r3, μ̄ = 2, 1.5, 0.5, 0.5, 0.5, 50
    Penalty schedule and threshold constants; chosen by hand within theory ranges, not fitted to data.
assumptions (7)
  • domain assumption Assumption 4.1: LICQ holds on F for problem (P_ρ)
    Ensures uniqueness of multipliers and nonsingularity of linear systems; standard in constrained optimization.
  • domain assumption Assumption 4.4: no nonnegative multipliers cause a zero gradient combination at infeasible points
    Used in Lemma 4.1 to show ρ_k stabilizes after finitely many updates; weaker than LICQ but not automatic.
  • domain assumption Assumption 4.3: H_k uniformly positive definite with bounds a1, a2
    Controls descent and boundedness of A_k^{-1}; standard quasi-Newton condition.
  • domain assumption Assumption 5.1: strict complementarity
    Required for local fast convergence analysis, standard in SQP literature.
  • domain assumption Assumption 5.2: SOSC
    Required for superlinear convergence, standard.
  • domain assumption Assumption 5.4: retraction is second-order
    Needed in Theorem 5.2 to prove unit step acceptance; restricts the class of admissible retractions.
  • domain assumption Assumption 5.5: Dennis-Moré condition on H_k
    Requires the Hessian approximation to be accurate along η_k; necessary for superlinear convergence.

how reviews work

0 comments
Cite this review

Pith. "Pith review of A globally and superlinearly convergent QO-free method for nonlinear optimization on Riemannian manifolds." pith.science (2026). https://pith.science/paper/36RARAYA

@misc{pith2026260809755,
  author       = {Pith},
  title        = {Pith review of: A globally and superlinearly convergent QO-free method for nonlinear optimization on Riemannian manifolds},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/36RARAYA}},
  note         = {Machine review of arXiv:2608.09755}
}
read the original abstract

The quadratic optimization-free (QO-free) method is a class of powerful and effective algorithms for solving nonlinearly constrained optimization problems in Euclidean spaces. The aim of the present work is to extend this method to solve optimization problems on manifolds with additional equality and inequality constraints. We first present a specific algorithm in the manifold setting. At each iteration, three linear systems sharing a common linear operator are solved to determine the master search direction. In addition, a higher-order correction direction is obtained by solving a reduced linear least squares subproblem to circumvent the Maratos effect which is assumed not to arise in existing related literature. A Riemannian arc search is then performed within the tangent space of the current iterate to generate the new iterate. Under appropriate assumptions, we establish the global and strong convergence of the proposed method. Moreover, we prove that the unit step size will eventually be accepted by the arc search, upon which the superlinear convergence of the algorithm is established. Finally, numerical results demonstrate that the proposed method is very competitive compared with other existing approaches.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

66 extracted references · 40 canonical work pages

  1. [14]

    SIAM Jou rnal on Opti- mization 11(1), 113–132 (2000) https://doi.org/10.1137/S1052623499353935

    Qi, H.D., Qi, L.Q.: A new QP-free, globally convergent, locally superlin early con- vergent algorithm for inequality constrained optimization. SIAM Jou rnal on Opti- mization 11(1), 113–132 (2000) https://doi.org/10.1137/S1052623499353935

    work page doi:10.1137/s1052623499353935 2000

  2. [1]

    IEEE Transactions on I nformation Theory 62(3), 1458–1484 (2015) https://doi.org/10.1109/TIT.2015.2457942

    Montanari, A., Richard, E.: Non-negative principal component an alysis: Message passing algorithms and sharp asymptotics. IEEE Transactions on I nformation Theory 62(3), 1458–1484 (2015) https://doi.org/10.1109/TIT.2015.2457942

    arXiv 2015

  3. [2]

    IEEE Transactions on Robotics 34(5), 1252–1265 (2018) https://doi.org/10.1109/TRO.2018.2830390

    Brossette, S., Escande, A., Kheddar, A.: Multicontact posture s computa- tion on manifolds. IEEE Transactions on Robotics 34(5), 1252–1265 (2018) https://doi.org/10.1109/TRO.2018.2830390

    arXiv 2018

  4. [3]

    arXiv pre print (2025) arXiv:2505.02323

    Teng, S., Lin, T.Y., Clark, W.A., Vasudevan, R., Ghaffari, M.: Riemannia n direct trajectory optimization of rigid bodies on matrix lie groups. arXiv pre print (2025) arXiv:2505.02323

    arXiv 2025

  5. [4]

    In: 2017 International Conference on Samplin g The- ory and Applications (SampTA), pp

    Carson, T., Mixon, D.G., Villar, S., Ward, R.: Manifold optimization for k-means clustering. In: 2017 International Conference on Samplin g The- ory and Applications (SampTA), pp. 73–77. IEEE, Tallinn, Estonia (2 017). 32 https://doi.org/10.1109/SampTA.2017.8024388

    work page arXiv 2017

  6. [5]

    Advances in Neural Information Processing Systems 36 (2024) https://doi.org/10.52202/075280-3504

    Jiang, B., Liu, Y.F.: A Riemannian exponential augmented Lagrangia n method for computing the projection robust Wasserstein distanc e. Advances in Neural Information Processing Systems 36 (2024) https://doi.org/10.52202/075280-3504

    work page doi:10.52202/075280-3504 2024

  7. [6]

    Athena Scient ific, Belmont, MA (2016)

    Bertsekas, D.P.: Nonlinear Programming, 3rd edn. Athena Scient ific, Belmont, MA (2016)

    work page 2016

  8. [7]

    In : Mixed Inte- ger Nonlinear Programming vol

    Gill, P.E., Wong, E.: Sequential quadratic programming methods. In : Mixed Inte- ger Nonlinear Programming vol. 154, pp. 147–224. Springer, New Yo rk (2011). https://doi.org/10.1007/978-1-4614-1927-3 6

    doi:10.1007/978-1-4614-1927-3 2011

Show all 66 references
  1. [8]

    SIAM Journal on Optimization 11(4), 1092– 1118 (2001) https://doi.org/10.1137/S1052623498344562

    Lawrence, C.T., Tits, A.L.: A computationally efficient feasible seque ntial quadratic programming algorithm. SIAM Journal on Optimization 11(4), 1092– 1118 (2001) https://doi.org/10.1137/S1052623498344562

    2001 doi

  2. [9]

    Optimization 71(6), 1603–1635 (2022) https://doi.org/10.1080/02331934.2020.1827406

    Jian, J.B., Tang, C.M., Hu, Q.J., Han, D.L.: A sequential quadratic pro gram- ming algorithm without a penalty function, a filter or a constraint qua lification for inequality constrained optimization. Optimization 71(6), 1603–1635 (2022) https://doi.org/10.1080/02331934.2020.1827406

    2022

  3. [10]

    Journal of Scientific Computing 102(2), 33 (2025) https://doi.org/10.1007/s10915-024-02743-7

    Jian, J.B., Chen, W.R., Tang, C.M., Yin, J.H.: A partially feasible Jacobi- type distributed SQO method for two-block general linearly constr ained smooth optimization. Journal of Scientific Computing 102(2), 33 (2025) https://doi.org/10.1007/s10915-024-02743-7

    2025 doi

  4. [11]

    SIAM Journal on Control and Optimization 26(4), 788–811 (1988) https://doi.org/10.1137/0326046

    Panier, E.R., Tits, A.L., Herskovits, J.N.: A QP-free, globally conve rgent, locally superlinearly convergent algorithm for inequality constrained optimiza- tion. SIAM Journal on Control and Optimization 26(4), 788–811 (1988) https://doi.org/10.1137/0326046

    1988 doi

  5. [12]

    Journal of Optimization Theory and Applications 95, 371–397 (1997) https://doi.org/10.1023/A:1022639306130

    Gao, Z.Y., He, G.P., Wu, F.: Sequential systems of linear equations algorithm for nonlinear optimization problems with general constrain ts. Journal of Optimization Theory and Applications 95, 371–397 (1997) https://doi.org/10.1023/A:1022639306130

    1997 doi

  6. [13]

    Mathematical Programming 85(1), 81–106 (1999) https://doi.org/10.1007/s101070050047

    Kanzow, C., Qi, H.D.: A QP-free constrained Newton-type metho d for vari- ational inequality problems. Mathematical Programming 85(1), 81–106 (1999) https://doi.org/10.1007/s101070050047

    1999 doi

  7. [15]

    SIAM Journal on Optimization 13(4), 1222– 1244 (2003) https://doi.org/10.1137/S1052623401383881

    Yang, Y.F., Li, D.H., Qi, L.Q.: A feasible sequential linear equation met hod for 33 inequality constrained optimization. SIAM Journal on Optimization 13(4), 1222– 1244 (2003) https://doi.org/10.1137/S1052623401383881

    2003 doi

  8. [16]

    SIAM Journal on Optimization 17(2), 401–429 (2006) https://doi.org/10.1137/040605904

    Chen, L.F., Wang, Y.L., He, G.P.: A feasible active set QP-free meth od for nonlinear programming. SIAM Journal on Optimization 17(2), 401–429 (2006) https://doi.org/10.1137/040605904

    2006 doi

  9. [17]

    Acta Mathematica Sinica, English Series 26(12), 2399–2420 (2010) https://doi.org/10.1007/s10114-010-7432-0

    Jian, J.B., Han, D.L., Xu, Q.J.: A new sequential systems of linear equations algorithm of feasible descent for inequality constrained o ptimiza- tion. Acta Mathematica Sinica, English Series 26(12), 2399–2420 (2010) https://doi.org/10.1007/s10114-010-7432-0

    2010 doi

  10. [18]

    Journal of Optimization The ory and Applications 113(2), 297–323 (2002) https://doi.org/10.1023/A:1014882909302

    Qi, L.Q., Yang, Y.F.: Globally and superlinearly convergent QP-free algorithm for nonlinear constrained optimization. Journal of Optimization The ory and Applications 113(2), 297–323 (2002) https://doi.org/10.1023/A:1014882909302

    2002 doi

  11. [19]

    Journal of Optimization Theory and Applications 119(2), 281–316 (2003) https://doi.org/10.1023/B:JOTA.0000005447.36961.29

    Facchinei, F., Lazzari, C.: Local feasible QP-free algorithms for the constrained minimization of SC 1 functions. Journal of Optimization Theory and Applications 119(2), 281–316 (2003) https://doi.org/10.1023/B:JOTA.0000005447.36961.29

    2003

  12. [20]

    Optimization Methods & Software 29(6), 1238–1260 (2014) https://doi.org/10.1080/10556788.2013.879587

    Liu, W.A., Shen, C.G., Zhu, X.J., Pu, D.G.: An infeasible QP-free algo- rithm without a penalty function or a filter for nonlinear inequality-co nstrained optimization. Optimization Methods & Software 29(6), 1238–1260 (2014) https://doi.org/10.1080/10556788.2013.879587

    2014

  13. [21]

    Springer, New Y ork (2006)

    Nocedal, J., Wright, S.J.: Numerical Optimization. Springer, New Y ork (2006)

    2006

  14. [22]

    Balkan Journal of Geometry and Its Applications 12(2) (2006)

    Munier, J.: Steepest descent method on a Riemannian manifold: t he convex case. Balkan Journal of Geometry and Its Applications 12(2) (2006). HAL ID: hal- 00018758

    2006

  15. [23]

    Optimization 64(4), 1011–1031 (2015) https://doi.org/10.1080/02331934.2013.836650

    Sato, H., Iwai, T.: A new, globally convergent Riemannian con- jugate gradient method. Optimization 64(4), 1011–1031 (2015) https://doi.org/10.1080/02331934.2013.836650

    2015

  16. [24]

    Computational Optimization and Applications 67(1), 73–110 (2017) https://doi.org/10.1007/s10589-016-9883-4

    Zhu, X.J.: A Riemannian conjugate gradient method for optimizat ion on the Stiefel manifold. Computational Optimization and Applications 67(1), 73–110 (2017) https://doi.org/10.1007/s10589-016-9883-4

    2017 doi

  17. [25]

    Journal of Optimization Theory and Applications 190(1), 130–150 (2021) https://doi.org/10.1007/s10957-021-01874-3

    Sakai, H., Iiduka, H.: Sufficient descent Riemannian conjugate gr adient meth- ods. Journal of Optimization Theory and Applications 190(1), 130–150 (2021) https://doi.org/10.1007/s10957-021-01874-3

    2021 doi

  18. [26]

    Numerical Algorithms 94(1), 131–147 (2023) https://doi.org/10.1007/s11075-022-01495-5 34

    Tang, C.M., Tan, W.C., Xing, S.J., Zheng, H.Y.: A class of spectral co njugate gradient methods for Riemannian optimization. Numerical Algorithms 94(1), 131–147 (2023) https://doi.org/10.1007/s11075-022-01495-5 34

    2023 doi

  19. [27]

    IMA Journal of Numerical Analysis 22(3), 359–390 (2002) https://doi.org/10.1093/imanum/22.3.359

    Adler, R.L., Dedieu, J.P., Margulies, J.Y., Martens, M., Shub, M.: New - ton’s method on Riemannian manifolds and a geometric model for the human spine. IMA Journal of Numerical Analysis 22(3), 359–390 (2002) https://doi.org/10.1093/imanum/22.3.359

    2002 doi

  20. [28]

    SIAM Journal on Optimization 25(3), 1660–1685 (2015) https://doi.org/10.1137/140955483

    Huang, W., Gallivan, K.A., Absil, P.A.: A Broyden class of quasi-Newto n methods for Riemannian optimization. SIAM Journal on Optimization 25(3), 1660–1685 (2015) https://doi.org/10.1137/140955483

    2015 doi

  21. [29]

    Numerical Algorithms 72(1), 57–90 (2016) https://doi.org/10.1007/s11075-015-0034-2

    Yousefpour, R.: Combination of steepest descent and BFGS me thods for non- convex nonsmooth optimization. Numerical Algorithms 72(1), 57–90 (2016) https://doi.org/10.1007/s11075-015-0034-2

    2016 doi

  22. [30]

    Foundations of Computational Mathematics 7(3), 303–330 (2007) https://doi.org/10.1007/s10208-005-0179-9

    Absil, P.A., Baker, C.G., Gallivan, K.A.: Trust-region methods on Riem annian manifolds. Foundations of Computational Mathematics 7(3), 303–330 (2007) https://doi.org/10.1007/s10208-005-0179-9

    2007 doi

  23. [31]

    IMA Journal of Numerical Analys is 36(3), 1167–1192 (2016) https://doi.org/10.1093/imanum/drv043

    Grohs, P., Hosseini, S.: Nonsmooth trust region algorithms for lo cally Lipschitz functions on Riemannian manifolds. IMA Journal of Numerical Analys is 36(3), 1167–1192 (2016) https://doi.org/10.1093/imanum/drv043

    2016 doi

  24. [32]

    Princeton University Press, Princeton, NJ (2008)

    Absil, P.A., Mahony, R., Sepulchre, R.: Optimization Algorithms on Ma trix Manifolds. Princeton University Press, Princeton, NJ (2008)

    2008

  25. [33]

    Cambridge University Press, Cambridge (2023)

    Boumal, N.: An Introduction to Optimization on Smooth Manifolds. Cambridge University Press, Cambridge (2023)

    2023

  26. [34]

    Springer, Switzerland (2021)

    Sato, H.: Riemannian Optimization and Its Applications. Springer, Switzerland (2021)

    2021

  27. [35]

    Journal of the Operations Research Society of China 8, 199–248 (2020) https://doi.org/10.1007/s40305-020-00295-9

    Hu, J., Liu, X., Wen, Z.W., Yuan, Y.X.: A brief introduction to manifold opti- mization. Journal of the Operations Research Society of China 8, 199–248 (2020) https://doi.org/10.1007/s40305-020-00295-9

    2020 doi

  28. [36]

    Pacific Journal of Optimiz ation 10(2), 415–434 (2014)

    Yang, W.H., Zhang, L.H., Song, R.: Optimality conditions for the non linear pro- gramming problems on Riemannian manifolds. Pacific Journal of Optimiz ation 10(2), 415–434 (2014). Zbl 1322.90096

    2014

  29. [37]

    SIAM Journal on Optimization 29(4), 2423– 2444 (2019) https://doi.org/10.1137/18M1181602

    Bergmann, R., Herzog, R.: Intrinsic formulation of KKT condition s and constraint qualifications on smooth manifolds. SIAM Journal on Optimization 29(4), 2423– 2444 (2019) https://doi.org/10.1137/18M1181602

    2019 doi

  30. [38]

    Applied Mathematics & Optimization 82(3), 949–981 (2020) https://doi.org/10.1007/s00245-019-09564-3

    Liu, C.S., Boumal, N.: Simple algorithms for optimization on Riemannian man- ifolds with constraints. Applied Mathematics & Optimization 82(3), 949–981 (2020) https://doi.org/10.1007/s00245-019-09564-3

    2020 doi

  31. [39]

    SIAM Journal on Optimization 31(3), 2255–2284 (2021) https://doi.org/10.1137/20M1341325

    Schiela, A., Ortiz, J.: An SQP method for equality constrained opt imization 35 on Hilbert manifolds. SIAM Journal on Optimization 31(3), 2255–2284 (2021) https://doi.org/10.1137/20M1341325

    2021 doi

  32. [40]

    SIAM Journa l on Optimization 32(2), 822–853 (2022) https://doi.org/10.1137/20M1370173

    Obara, M., Okuno, T., Takeda, A.: Sequential quadratic optimiza tion for nonlinear optimization problems on Riemannian manifolds. SIAM Journa l on Optimization 32(2), 822–853 (2022) https://doi.org/10.1137/20M1370173

    2022 doi

  33. [41]

    Journal of Optimization Theory and Applications 22(3), 297–309 (1977) https://doi.org/10.1007/BF00932858

    Han, S.P.: A globally convergent method for nonlinear programmin g. Journal of Optimization Theory and Applications 22(3), 297–309 (1977) https://doi.org/10.1007/BF00932858

    1977 doi

  34. [42]

    PhD thesis, Imperial College London (Un iversity of London) (1978)

    Maratos, N.: Exact penalty function algorithms for finite dimens ional and con- trol optimization problems. PhD thesis, Imperial College London (Un iversity of London) (1978)

    1978

  35. [43]

    IEEE Transactions on Automatic Control 69(3), 2060–2066 (2024) https://doi.org/10.1109/TAC.2023.3318195

    Obara, M., Sato, K., Sakamoto, H., Takayuki, O., Akiko, T.: Stable linear system identification with prior knowledge by Riemannian sequential quadrat ic opti- mization. IEEE Transactions on Automatic Control 69(3), 2060–2066 (2024) https://doi.org/10.1109/TAC.2023.3318195

    2024

  36. [44]

    Journal of Optimization Theory and Application s, 1–37 (2024) https://doi.org/10.1007/s10957-024-02403-8

    Lai, Z.J., Yoshise, A.: Riemannian interior point methods for const rained opti- mization on manifolds. Journal of Optimization Theory and Application s, 1–37 (2024) https://doi.org/10.1007/s10957-024-02403-8

    2024 doi

  37. [45]

    arXiv preprint (2025) arXiv:2505.19724

    Obara, M., Okuno, T., Takeda, A.: Local near-quadratic conve rgence of Rieman- nian interior point methods. arXiv preprint (2025) arXiv:2505.19724

    2025 arXiv

  38. [46]

    arXiv preprint (2025) arXiv:2501.15419

    Obara, M., Okuno, T., Takeda, A.: A primal-dual interior point tru st region method for second-order stationary points of Riemannian inequalit y-constrained optimization problems. arXiv preprint (2025) arXiv:2501.15419

    2025 arXiv

  39. [47]

    Mathematical Programming 11, 67–80 (1976) https://doi.org/10.1007/BF01580371

    Mayne, D.Q., Polak, E.: Feasible directions algorithms for optimizat ion problems with equality and inequality constraints. Mathematical Programming 11, 67–80 (1976) https://doi.org/10.1007/BF01580371

    1976 doi

  40. [48]

    SIAM Journal on Optimization 14(1), 173–199 (2003) https://doi.org/10.1137/S1052623401392123

    Tits, A.L., W¨ achter, A., Bakhtiari, S., Urban, T.J., Lawrence, C.T.: A primal-dual interior-point method for nonlinear programming with strong global and local convergence properties. SIAM Journal on Optimization 14(1), 173–199 (2003) https://doi.org/10.1137/S1052623401392123

    2003 doi

  41. [49]

    Optimiz ation 24(3-4), 269–284 (1992) https://doi.org/10.1080/02331939208843795

    Fischer, A.: A special Newton-type optimization method. Optimiz ation 24(3-4), 269–284 (1992) https://doi.org/10.1080/02331939208843795

    1992 doi

  42. [50]

    In: Recent Advances in Nonsmooth Optimization, pp

    Fischer, A.: An NCP–function and its use for the solution of comp lementarity problems. In: Recent Advances in Nonsmooth Optimization, pp. 88– 105. World Scientific, Singapore (1995). https://doi.org/10.1142/9789812812827 0006 36

    1995 doi

  43. [51]

    Berlin-New York (1982)

    Klingenberg, W.: Riemannian geometry. Berlin-New York (1982)

    1982

  44. [52]

    SIA M Journal on Scientific and Statistical Somputing 4(3), 553–572 (1983) https://doi.org/10.1137/0904038

    Mor´ e, J.J., Sorensen, D.C.: Computing a trust region step. SIA M Journal on Scientific and Statistical Somputing 4(3), 553–572 (1983) https://doi.org/10.1137/0904038

    1983 doi

  45. [53]

    Mathem atics of Opera- tions Research 5(1), 43–62 (1980) https://doi.org/10.1287/moor.5.1.43

    Robinson, S.M.: Strongly regular generalized equations. Mathem atics of Opera- tions Research 5(1), 43–62 (1980) https://doi.org/10.1287/moor.5.1.43

    1980 doi

  46. [54]

    Mathema tical Program- ming 198(1), 855–897 (2023) https://doi.org/10.1007/s10107-022-01794-8

    Jiang, B., Meng, X., Wen, Z.W., Chen, X.J.: An exact penalty approa ch for optimization with nonnegative orthogonality constraints. Mathema tical Program- ming 198(1), 855–897 (2023) https://doi.org/10.1007/s10107-022-01794-8

    2023 doi

  47. [55]

    The Journal of Machine Learning Rese arch 15(1), 1455–1459 (2014)

    Boumal, N., Mishra, B., Absil, P.A., Sepulchre, R.: Manopt, a Matlab t oolbox for optimization on manifolds. The Journal of Machine Learning Rese arch 15(1), 1455–1459 (2014)

    2014

  48. [56]

    Advances in Computational Mathematics 46, 1–25 (2020) https://doi.org/10.1007/s10444-020-09779-x

    Guglielmi, N., Scalone, C.: An efficient method for non-negative low- rank completion. Advances in Computational Mathematics 46, 1–25 (2020) https://doi.org/10.1007/s10444-020-09779-x

    2020 doi

  49. [57]

    SIAM Journal on Optimization 23(2), 1214–1236 (2013) https://doi.org/10.1137/110845768

    Vandereycken, B.: Low-rank matrix completion by Riemannian op ti- mization. SIAM Journal on Optimization 23(2), 1214–1236 (2013) https://doi.org/10.1137/110845768

    2013 doi

  50. [58]

    Numerische Mathemat ik 136(2), 523– 543 (2017) https://doi.org/10.1007/s00211-016-0848-4

    Huang, W., Absil, P.A., Gallivan, K.A.: Intrinsic representation of ta ngent vectors and vector transports on matrix manifolds. Numerische Mathemat ik 136(2), 523– 543 (2017) https://doi.org/10.1007/s00211-016-0848-4

    2017 doi

  51. [59]

    European Journal of Operational Research 200(3), 645–657 (2010) https://doi.org/10.1016/j.ejor.2009.01.052

    Jian, J.B., Tang, C.M., Zheng, H.Y.: Sequential quadratically const rained quadratic programming norm-relaxed algorithm of strongly sub-fe asible direc- tions. European Journal of Operational Research 200(3), 645–657 (2010) https://doi.org/10.1016/j.ejor.2009.01.052

    2010 doi

  52. [60]

    Tang, C.M., Jian, J.B.: Strongly sub-feasible direction method for constrained optimization problems with nonsmooth objective func- tions. European Journal of Operational Research 218(1), 28–37 (2012) https://doi.org/10.1016/j.ejor.2011.10.055 Appendix A The Proofs in Section ...

    2012 doi

  53. [61]

    Thus, the proof is complete

    that ¯λ = 0. Thus, the proof is complete. □ Proof of Lemma 3.2 If ηk1 and ηk have been generated by Algorithm 1, then we can conclude that ηk0 ⁄= 0 xk . In fact, if ηk0 = 0 xk , then it follows from ( 10b) that√ 2βk i λk0 i = 0 for all i ∈ L . This along with ( 11) shows that ...

  54. [62]

    where the last inequality is due to ( 13)

    and ( 63) shows that ⟨gradFρk (xk), ηk1⟩ = ⟨gradFρk (xk), ηk0⟩ − ∑ i:λk0 i ⁄=0 λk0 i (min{λk0 i , 0})3 = ⟨gradFρk (xk), ηk0⟩ − ∑ i:λk0 i <0 (λk0 i )4 < 0. where the last inequality is due to ( 13). (ii) The definitions of ηk and θk show that ⟨gradFρk (xk), ηk⟩ = ⟨gradFρk (xk), ...

  55. [66]

    (72) Similarly, for i /∈ I (x∗), we obtain { λk0 i zk } K′ → ˆλ∗ i = 0, i / ∈ I (x∗)

    and ( 70), we obtain ˆλ∗ i ≥ −w, ∀ i ∈ E ; ˆλ∗ I ≥ 0. (72) Similarly, for i /∈ I (x∗), we obtain { λk0 i zk } K′ → ˆλ∗ i = 0, i / ∈ I (x∗). Finally, dividing both sides of (

  56. [67]

    (73) Since ˆλ∗ E and ˆλ∗ I are not both zero, it follows from Assumption 4.1 that E(x∗) ⁄= E (i.e., x∗ ∈ F \FP ) and w > 0

    by zk, taking the limit over K′, and combining (64), Assumption 4.3, and the twice continuous differentiability of f (x), we obtain ∑ i∈E ˆλ∗ i gradci(x∗) + ∑ i∈I(x∗) ˆλ∗ i gradci(x∗) = 0 x∗. (73) Since ˆλ∗ E and ˆλ∗ I are not both zero, it follows from Assumption 4.1 that E(x∗...

  57. [68]

    For i /∈ E (x∗), we have β∗ i > 0, and thus { λk0 i zk } K′ → 0, ∀i /∈ E (x∗)

    by zk and taking the limit over K′, from the boundedness of {αk i }K′ and ( 64), we have { βk i λk0 i zk } K′ → 0. For i /∈ E (x∗), we have β∗ i > 0, and thus { λk0 i zk } K′ → 0, ∀i /∈ E (x∗). It follows from ( 70) that there exists a constant w ≥ 0 such that {ρk/zk}K′ → w an...

  58. [75]

    Summarizing the above analysis, we can conclude that ( x∗, λ∗) is a KKT pair of problem ( Pρ)

    shows that λ∗ i = 0 for all i /∈ L(x∗). Summarizing the above analysis, we can conclude that ( x∗, λ∗) is a KKT pair of problem ( Pρ). Next, we show that {λk0}K → λ∗. In fact, if there is another accumu- lation point of {λk0}K, denoted by ¯λ∗. From the above analysis, we can i...

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.