Abstract
Canonical duality theory (CDT) is presented by its creator DY Gao as a theory which can be used for solving a large class of challenging real-world problems. It is the aim of this paper to study rigorously constrained optimization problems in finite dimensional spaces using the method suggested by CDT and to discuss several results published in the last ten years.
Similar content being viewed by others
Notes
The reference “[7]” is “Gao, D.Y.: Nonconvex semi-linear problems and canonical duality solutions, in Advances in Mechanics and Mathematics II. In: Gao, D.Y., Ogden R.W. (eds.), pp. 261–311. Kluwer Academic Publishers (2003)”.
The reference “[26]” mentioned in [14, Rem. 1] is the item [7] from our bibliography, the others being the following: “8. Fang, S.C., Gao, D.Y., Sheu, R.L., Wu, S.Y.: Canonical dual approach to solving 0–1 quadratic programming problems. J. Ind. Manag. Optim. 4(4), 125–142 (2008)”, “12. Gao, D.Y.: Solutions and optimality criteria to box constrained nonconvex minimization problem. J. Ind. Manag. Optim. 3(2), 293–304 (2007)”, and “16. Gao, D.Y., Ruan, N.: Solutions to quadratic minimization problems with box and integer constraints. J. Glob. Optim. 47, 463–484 (2010)”, respectively.
The emphasized text can be found also in [3, p. NP26] and [4, p. 33]. One must also observe that for Latorre and Gao \(\mu \ne 0\) is equivalent to “\(\mu _{j}\ne 0\) \(\forall j=1,...,p\)”, and \((\lambda ,\mu )\in {\mathbb {R}}^{m\times p}\) if \(\lambda \in {\mathbb {R}}^{m}\) and \(\mu \in {\mathbb {R}}^{p}.\)
It is worth quoting DY Gao’s comment from [1, p. 19] on our remark from [17, p. 1783] that the proof of [7, Th. 2] is not convincing: “Regarding the so-called “not convincing proof”, serious researcher should provide either a convincing proof or a disproof, rather than a complaint. Note that the canonical dual variables \(\sigma _{0}\) and \(\sigma _{1}\) are in two different levers (scales) with totally different physical units\(^{14}\), it is completely wrong to consider \((\sigma _{0},\sigma _{1})\) as one vector and to discuss the concavity of \(\varXi _{1}\left( x,(\cdot ,\cdot )\right) \) on \({\mathscr {S}}_{a} ^{+}\). The condition “\({\mathscr {S}}_{a}^{+}\) is convex” in Theorem 2 [5] should be understood in the way that \({\mathscr {S}}_{a}^{+}\) is convex in \(\sigma _{0}\) and \(\sigma _{1}\), respectively, as emphasized in Remark 1 [5]. Thus, the proof of Theorem 2 given in [5] is indeed convincing by simply using the classical saddle min-max duality for \((x,\sigma _{0})\) and \((x,\sigma _{1})\), respectively.”
Note 14 from the text above is “Let us consider Example 1 in [5]. If the unit for x is the meter (m) and for q is Kg/m, then the units for the Lagrange multiplier \(\mu \) (dual to the constraint \(g(x)=\tfrac{1}{2}(\tfrac{1}{2}x^{2}-d)^{2}-e\)) should be \(Kg/m^{3}\) and for \(\sigma \) (canonical dual to \(\varLambda (x)=\tfrac{1}{2}x^{2}\)) should be Kg/m, respectively, so that each terms in \(\varXi _{1}(x,\mu ,\sigma )\) make physical sense”; “[5]” is our reference [7].
In fact, we did not find a correct proof of the “min-max” duality (like Eq. (20) from [10]) in DY Gao’s papers in the case \(Q_{0}\ne \overline{1,m}\).
Referring to [9, Lem. 1], which is a reformulation of [15, Assertion II], the authors say: “The following Lemma is well known in mathematical programming (cf. Latorre and Gao [12] and Voisei and Zălinescu [13])”; the references “[12]” and “[13]” are our items [7] and [15], respectively. Of course, [9, Lem. 1] isFootnote 7 (continued)
not “well known in mathematical programming”, being very specific to the problem considered in [6]. The reference [7] is not mentioned in [10] with respect to [9, Lem. 1].
The reference “[3]” is the item [6] from our bibliography.
The text is quoted from [10, p. 370]; here the reference “[15]” is “Wu, C., Gao, D.Y.: Canonical primal-dual method for solving nonconvex minimization problems. In: Gao, D.Y., Latorre, V., Ruan, N. (eds.) Advances in Canonical Duality Theory. Springer, Berlin”. Note that the same text can be found in [8, p. 9] without any reference, as well as in [9, p. NP236], where the indicated reference is “Wu, C, Li, C, and Gao, DY. Canonical primal-dual method for solving nonconvex minimization problems. arXiv:1212.6492, 2012.” Observe that the main difference between arXiv:1212.6492 and reference “[15]” of [10] consists in the list of the authors, the content being practically the same.
References
Gao, D.Y.: On unified modeling, canonical duality-triality theory, challenges and breakthrough in optimization, arXiv:1605.05534v3 (2016)
Gao, D.Y., Latorre, V., Ruan, N.: Canonical Duality Theory Unified Methodology for Multidisciplinary Study. Advances in Mechanics and Mathematics, vol. 37. Springer, Cham (2017)
Gao, D.Y., Ruan, N., Latorre, V.: RETRACTED: Canonical duality-triality theory: bridge between nonconvex analysis/mechanics and global optimization in complex system, Math. Mech. Solids, 21(3), NP5–NP36 (2016) (see also arXiv:1410.2665)
Gao, D.Y., Ruan, N., Latorre, V.: Canonical duality-triality theory: bridge between nonconvex analysis/mechanics and global optimization in complex system. In: Gao, D.Y., Latorre, V., Ruan, N. (eds.) Canonical Duality Theory. Advances in Mechanics and Mathematics, pp. 1–47. Springer, Cham (2017)
Gao, D.Y., Ruan, N., Sherali, H.: Solutions and optimality criteria for nonconvex constrained global optimization problems with connections between canonical and Lagrangian duality. J. Global Optim. 45, 473–497 (2009)
Gao, D.Y., Yang, W.C.: Complete solutions to minimal distance problem between two nonconvex surfaces. Optimization 57, 705–714 (2008)
Latorre, V., Gao, D.Y.: Canonical duality for solving general nonconvex constrained problems. Optim. Lett. 10, 1763–1779 (2016)
Morales-Silva, D., Gao, D.Y.: On the minimal distance between two surfaces, arXiv:1210.1618 (2012) (compare with [9] and [10])
Morales-Silva, D., Gao, D.Y.: RETRACTED: on the minimal distance between two non-convex surfaces. Math. Mech. Solids 21(3), NP225–NP237 (2016). (compare with [8] and [10])
Morales-Silva, D., Gao, D.Y.: On minimal distance between two surfaces. In: Gao, D.Y., Latorre, V., Ruan, N. (eds.) Canonical Duality Theory Advances in Mechanics and Mathematics, pp. 359–371. Springer, Cham (2017). (compare with [8] and [9])
Rockafellar, R.T.: Convex Analysis. Princeton University Press, New York (1972)
Ruan, N., Gao, D.Y.: RETRACTED: canonical duality theory for solving nonconvex/discrete constrained global optimization problems. Math. Mech. Solids 21(3), NP194–NP205 (2016)
Ruan, N., Gao, D.Y.: Canonical duality theory for solving nonconvex/discrete constrained global optimization problems. In: Gao, D.Y., Latorre, V., Ruan, N. (eds.) Canonical Duality Theory Advances in Mechanics and Mathematics, pp. 187–201. Springer, Cham (2017)
Ruan, N., Gao, D.Y.: Global optimal solution to quadratic discrete programming problem with inequality constraints. In: Latorre, V., Ruan, N. (eds.) Canonical Duality Theory Advances in Mechanics and Mathematics, pp. 315–338. Springer, Cham (2017)
Voisei, M.D., Zălinescu, C.: A counter-example to ‘minimal distance between two non-convex surfaces’. Optimization 60, 593–602 (2011)
Voisei, M.D., Zălinescu, C.: Counterexamples to some triality and tri-duality results. J. Global Optim. 49, 173–183 (2011)
Zălinescu, C.: On V. Latorre and D.Y. Gao’s paper “Canonical duality for solving general nonconvex constrained problems”. Optim. Lett. 10, 1781–1787 (2016)
Zălinescu, C.: On quadratic optimization problems and canonical duality theory. Optim. Lett. 14, 2227–2245 (2020)
Zălinescu, C.: On unconstrained optimization problems solved using the canonical duality and triality theories. Optimization 69, 2551–2576 (2020)
Zălinescu, C.: On constrained optimization problems solved using the canonical duality theory, In: Le Thi, H.A., Le, H.M., Dinh, T.P. (eds.) Optimization of Complex Systems: Theory, Models, Algorithms and Applications. WCGO 2019, Advances in Intelligent Systems and Computing, pp. 155–163. Springer, Cham (2020)
Acknowledgements
We thank Prof. Marius Durea for reading a previous version of the paper and for his useful remarks. This work was presented at the 6th World Congress on Global Optimization, Metz, 2019; see [20] for an abridged version which contains the main results without proofs.
Author information
Authors and Affiliations
Corresponding author
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
About this article
Cite this article
Zălinescu, C. On canonical duality theory and constrained optimization problems. J Glob Optim 82, 1053–1070 (2022). https://doi.org/10.1007/s10898-021-01021-2
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10898-021-01021-2
Keywords
- Constrained optimization problem
- Dual function
- Extended Lagrangian
- Canonical duality theory
- Counterexample