Abstract
In this paper, we introduce a geometric theory which relates a geometric structure of convex optimization problems to computational complexity to solve the problems. Specifically, we develop information geometric framework of conic linear optimization problems and show that the iteration complexity of the standard polynomial-time primal-dual predictor-corrector interior-point algorithms to solve symmetric cone programs is written with an information geometric curvature integral of the central path which the algorithm traces. Numerical experiments demonstrate that the number of iterations is quite well explained with the integral even for the large problems with thousands of variables; we claim that the iteration-complexity of the primal-dual predictor-corrector path-following algorithm is an information geometric quantity. We also develop a global theorem about the central path for linear programs.
This is a preview of subscription content, log in via an institution to check access.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Preview
Unable to display preview. Download preview PDF.
References
Amari, S.-I., Nagaoka, H.: Methods of information geometry. Translations of Mathematical Monographs, vol. 191, p. 206. American Mathematical Society, Providence (2000); Translated from the 1993 Japanese original by Daishi Harada
Faraut, J., Korányi, A.: Analysis on symmetric cones. Oxford Mathematical Monographs. The Clarendon Press Oxford University Press, Oxford Science Publications, New York (1994)
Kakihara, S., Ohara, A., Tsuchiya, T.: Curvature Integrals and Iteration Complexities in SDP and Symmetric Cone Programs. Optimization Online (2011)
Kakihara, S., Ohara, A., Tsuchiya, T.: Information Geometry and Interior-Point Algorithms in Semidefinite Programs and Symmetric Cone Programs. J. Optim. Theory Appl. 157(3), 749–780 (2013)
Karmarkar, N.: Riemannian geometry underlying interior-point methods for linear programming. In: Mathematical Developments Arising from Linear Programming (Brunswick, ME, 1988). Contemp. Math., vol. 114, pp. 51–75. Amer. Math. Soc., Providence (1990)
Monteiro, R.D.C., Tsuchiya, T.: A strong bound on the integral of the central path curvature and its relationship with the iteration-complexity of primal-dual path-following LP algorithms. Math. Program. 115(1, ser. A), 105–149 (2008)
Nesterov, Y., Nemirovskii, A.: Interior-point polynomial algorithms in convex programming. SIAM Studies in Applied Mathematics, vol. 13, p. 405. Society for Industrial and Applied Mathematics (SIAM), Philadelphia (1994)
Nesterov, Y.E., Todd, M.J.: On the Riemannian geometry defined by self-concordant barriers and interior-point methods. Found. Comput. Math. 2(4), 333–361 (2002)
Ohara, A., Suda, N., Amari, S.-I.: Dualistic differential geometry of positive definite matrices and its applications to related problems. Linear Algebra Appl. 247, 31–53 (1996)
Shima, H.: The geometry of Hessian structures. World Scientific, Hackensack (2007)
Tanabe, K.: Center flattening transformation and a centered newton method for linear programming. Manuscript Presented at MP Seminar, The Operations Research Society of Japan (1987)
Tanabe, K.: Centered newton methods and differential geometry of optimization. Cooperative Research Report, The Institute of Statistical Mathematics, 89 (1996)
Tanabe, K., Tsuchiya, T.: New geometry of linear programming. Suurikagaku 303, 32–37 (1988) (in Japanese)
Vavasis, S.A., Ye, Y.: A primal-dual interior point method whose running time depends only on the constraint matrix. Math. Programming 74(1, ser. A), 79–120 (1996)
Author information
Authors and Affiliations
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2013 Springer-Verlag Berlin Heidelberg
About this paper
Cite this paper
Kakihara, S., Ohara, A., Tsuchiya, T. (2013). Information Geometry and Interior-Point Algorithms. In: Nielsen, F., Barbaresco, F. (eds) Geometric Science of Information. GSI 2013. Lecture Notes in Computer Science, vol 8085. Springer, Berlin, Heidelberg. https://doi.org/10.1007/978-3-642-40020-9_87
Download citation
DOI: https://doi.org/10.1007/978-3-642-40020-9_87
Publisher Name: Springer, Berlin, Heidelberg
Print ISBN: 978-3-642-40019-3
Online ISBN: 978-3-642-40020-9
eBook Packages: Computer ScienceComputer Science (R0)