Editing Quadratic programming (section)

==Run-time complexity==

=== Convex quadratic programming ===
For [[positive-definite matrix|positive definite]] {{mvar|Q}}, when the problem is convex, the [[ellipsoid method]] solves the problem in (weakly) [[polynomial time]].<ref>{{cite journal| last=Kozlov | first=M. K. |author2=S. P. Tarasov | author3-link=Leonid Khachiyan |author3=Leonid G. Khachiyan | year=1979 | title=[Polynomial solvability of convex quadratic programming] | journal=[[Doklady Akademii Nauk SSSR]] | volume=248 | pages=1049–1051}} Translated in: {{cite journal| journal=Soviet Mathematics - Doklady | volume=20 | pages=1108–1111}}</ref>

Ye and Tse<ref>{{Cite journal |last1=Ye |first1=Yinyu |last2=Tse |first2=Edison |date=1989-05-01 |title=An extension of Karmarkar's projective algorithm for convex quadratic programming |url=https://doi.org/10.1007/BF01587086 |journal=Mathematical Programming |language=en |volume=44 |issue=1 |pages=157–179 |doi=10.1007/BF01587086 |s2cid=35753865 |issn=1436-4646|url-access=subscription }}</ref> present a polynomial-time algorithm, which extends [[Karmarkar's algorithm]] from linear programming to convex quadratic programming. On a system with ''n'' variables and ''L'' input bits, their algorithm requires O(L n) iterations, each of which can be done using O(L n<sup>3</sup>) arithmetic operations, for  a total runtime complexity of O(''L''<sup>2</sup> ''n''<sup>4</sup>). 

Kapoor and Vaidya<ref>{{Cite book |last1=Kapoor |first1=S |last2=Vaidya |first2=P M |chapter=Fast algorithms for convex quadratic programming and multicommodity flows |date=1986-11-01 |title=Proceedings of the eighteenth annual ACM symposium on Theory of computing - STOC '86 |chapter-url=https://dl.acm.org/doi/10.1145/12130.12145 |location=New York, NY, USA |publisher=Association for Computing Machinery |pages=147–159 |doi=10.1145/12130.12145 |isbn=978-0-89791-193-1|s2cid=18108815 }}</ref> present another algorithm, which requires O(''L'' * log ''L'' ''* n''<sup>3.67</sup> * log ''n'') arithmetic operations.

=== Non-convex quadratic programming ===
If {{mvar|Q}} is indefinite, (so the problem is non-convex) then the problem is [[NP-hard]].<ref>{{cite journal | last = Sahni | first = S. | title = Computationally related problems | journal = SIAM Journal on Computing | volume = 3 | issue = 4 | pages = 262–279 | year = 1974 | doi=10.1137/0203021| url = http://www.cise.ufl.edu/~sahni/papers/comp.pdf | citeseerx = 10.1.1.145.8685 }}</ref> A simple way to see this is to consider the non-convex quadratic constraint ''x<sub>i</sub>''<sup>2</sup> = ''x<sub>i</sub>''. This constraint is equivalent to requiring that ''x<sub>i</sub>'' is in {0,1}, that is, ''x<sub>i</sub>'' is a binary integer variable. Therefore, such constraints can be used to model any [[Integer programming|integer program]] with binary variables, which is known to be NP-hard.

Moreover, these non-convex problems might have several stationary points and local minima. In fact, even if {{mvar|Q}} has only one negative [[eigenvalue]], the problem is (strongly) [[NP-hard]].<ref>{{cite journal | title = Quadratic programming with one negative eigenvalue is (strongly) NP-hard | first1 = Panos M. | last1 = Pardalos | first2 = Stephen A. | last2 = Vavasis | journal = Journal of Global Optimization | volume = 1 | issue = 1 | year = 1991 | pages = 15–22 | doi=10.1007/bf00120662| s2cid = 12602885 }}</ref>

Moreover, finding a KKT point of a non-convex quadratic program is CLS-hard.<ref>{{Cite arXiv |eprint=2311.13738 |last1=Fearnley |first1=John |last2=Goldberg |first2=Paul W. |last3=Hollender |first3=Alexandros |last4=Savani |first4=Rahul |title=The Complexity of Computing KKT Solutions of Quadratic Programs |date=2023 |class=cs.CC }}</ref>