Share:

MTM2006-05550


Interior-point Methods for Large-Scale Optimization. Application to Statistical Data Protection.

Funding organization: Spanish Ministry of Science and Education. October 2006-September 2009.

Members and partners

Members
  • Jordi Castro
  • José Antonio González
  • Antonio Frangioni
  • Claudio Gentile
  • Jordi Cuesta
  • Adebayo Alabi
Universities
  • Universitat Politècnica de Catalunya (UPC)
  • Universita di Pisa
  • Consiglio Nazionale delle Ricerche, Roma, Italia
National Agencies
  • Statistics Germany
  • Statistics Netherlands
  • Statistics Italy
  • Statistics Catalonia
Top
 

Introduction and Objectives

The current information society requires continuous manage of  a large amount of data for decision making. Among the main agents devoted to the dissemination of aggregated or tabular data, we find  National Statistical Agencies (NSAs). However they must preserve the individual rights of privacy and data confidentiality. Data protection techniques are the tools used by NSAs for the safe dissemination of information. Several European NSAs are considering a new strategy for data protection: minimum-distance controlled tabular adjustment, which was suggested by this research group in a previous research project. This technique results  in a linear or quadratic optimization problem, depending on the distance to be used, which in practice turned out to be a challenge for nonpolynomial methods such as the simplex. Preliminary results with small- and medium-size tables showed that interior-point methods (in which the research group has vast experience) are more effective, being up to 60 times faster than the simplex in some instances. However, for real tables, with millions of cells and hundreds of thousands of constraints, general interior-point algorithms are simply inefficient and we need specialized ones. The objective of the project is the development of interior-point algorithms that are able to efficiently solve massive tabular data. For this purpose, the tasks, among others, are:

  • the constraint matrix structure of tabular data will be exploited;
  • iterative methods will be used for the linear systems at each interior-point iteration;
  • preconditioners for improving the convergence of iterative solvers will be used and developed
  • quadratic regularizations to the objective functions will be attempted to improve the preconditioners;
  • alternative solution procedures for the optimization problem will be studied (like recent self-regular interior-points, potential function approximation algorithms, heuristics, etc.).

Ultimately, we plan to develop a leading tool for statistical data protection, based on interior-point methods, that can be used by any NSA.
Top

Results

Publications

  • J. Castro, A. Frangioni, C. Gentile. Perspective reformulations of the CTA problem with L2 distances. Working paper.
  • J. Castro, A. Ouorou. An interior-point algorithm for routing in data telecommunications networks. Working paper.
  • J. Castro, Statistical disclosure control in tabular data, Research Report DR 2009/11, Dept. of Statistics and Operations Research, Universitat Politècnica de Catalunya, 2009. To appear as a chapter of Privacy and Anonymity in Information Management Systems, Springer, 2010.
  • J.A. González, J. Castro, Block coordinate descent decomposition for statistical data protection using controlled tabular adjustment, Research Report DR 2009/10, Dept. of Statistics and Operations Research, Universitat Politècnica de Catalunya, 2009. Submitted to Computers and Operations Research.
  • J. Castro, J. Cuesta, Improving an interior-point algorithm for multicommodity flows by quadratic regularizations, Research Report DR 2009/09, Dept. of Statistics and Operations Research, Universitat Politècnica de Catalunya, 2009. Submitted to Networks.
  • J. Cuesta, J. Castro, Existence and uniqueness of the regularized primal-dual central path, Research Report DR 2009/08, Dept. of Statistics and Operations Research, Universitat Politècnica de Catalunya, 2009. Submitted to Operations Research Letters (first revision in progress).
  • J. Castro, J. Cuesta, Quadratic regularizations in an interior-point method for primal block-angular problems, Mathematical Programming, DOI:10.1007/s10107-010-0341-2, in press (2010). Previously appeared as Research Report DR 2008/07, Dept. of Statistics and Operations Research, Universitat Politècnica de Catalunya, 2008.
  • A. Alabi, J. Castro, A mathematical model of refinery operations characterised by complete horizontal integration of subsystems from purchase to product distribution, Hydrocarbon World, 4(2) (2009) 55-56.
  • J. Castro, D. Baena, Using a mathematical programming modeling language for optimal CTA, Lecture Notes in Computer Science, 5262 (2008) 1-12, eds. J. Domingo-Ferrer and Y. Saigin, © Springer.
  • A. Alabi, J. Castro. Dantzig-Wolfe and block coordinate-descent decomposition in large-scale integrated refinery-planning, Computers and Operations Research, 36 (2009) 2472-2483. Previously appeared as Research Report DR 2008/01, Dept. of Statistics and Operations Research, Universitat Politècnica de Catalunya, 2008.
  • J. Castro, J.A. González, D. Baena. User's and programmer's manual of the RCTA package, Technical Report DR 2009/01, Dept. of Statistics and Operations Research, Universitat Politècnica de Catalunya, 2009.

 Top

Publications partially related with the project

  • D. Rosas, J. Castro and L. Montero, Using ACCPM in a simplicial decomposition algorithm for the traffic assignment problem, Computational Optimization and Applications, accepted in 2007, doi:10.1007/s10589-007-9153-6. Previously appeared as Research Report DR 2002/04, Dept. of Statistics and Operations Research, Universitat Politècnica de Catalunya.
  • J. Castro, A stochastic programming approach to cash management in banking, European Journal of Operational Research, 192 (2009) 963-974. Previously appeared as Research Report DR 2004/14, Dept. of Statistics and Operations Research, Universitat Politècnica de Catalunya, 2004.

Top

Conference and seminar presentations

  • J. Castro, J.A. González, A package for $L_1$ controlled tabular adjustment. Joint UNECE/Eurostat Work Session on Statistical Data Confidentiality, Bilbao (Basque Country), December 2009. Invited presentation.  
  • J. Castro, Advances on statistical disclosure control: tabular data protection of business data. IDESCAT seminar/technical session, IDESCAT, the National Statistical Institute of Catalonia, Barcelona, July 17 2009. Invited presentation.  
  • J. Castro, J. Cuesta, Improving an interior-point algorithm for multicommodity flows by quadratic regularizations, International Network Optimization Conference 2009, Pisa, Italia, April 2009. Invited presentation.
  • J. Castro, A. Ouorou, An interior-point algorithm for routing in data telecommunications networks, International Network Optimization Conference 2009, Pisa, Italia, April 2009. Invited presentation.  
  • J. Castro, J.A. González, Métodos heurísticos y exactos para el problema de ajuste controlado de tablas, XXXI Congreso Nacional de Estadística e Investigación Operativa, Murcia, Spain, February 2009.
  • J. Castro, J. Cuesta, Mejora de métodos iterativos en algoritmos de punto interior a través de regularizaciones cuadràticas, XXXI Congreso Nacional de Estadística e Investigación Operativa, Murcia, Spain, February 2009.  
  • J. Castro, A stochastic programming approach to cash management in banking, Seminar of the Faculty of Mathematics, Universidad de Sevilla, Sevilla (Spain), November 18, 2008. Invited presentation.
  • J. Castro, Improving iterative solvers in IPMs for structured problems through quadratic regularizations, Research seminar of the Department of Statistics, Universidad Carlos III de Madrid, Getafe (Spain), October 3, 2008. Invited presentation.  
  • J. Castro, D. Baena, Using a Mathematical Programming Modeling Language for Optimal CTA, Privacy in Statistical Databases 2008, Istanbul (Turkey), September 2008.  
  • J. Castro and J. Cuesta, Improving iterative solvers in IPMs for structured problems through quadratic regularizations, International Conference on Applied Mathematical Programming and Modelling APMOD, Bratislava (Slovak Republic), May 2008. Invited presentation.  
  • J. Castro, Iterative solvers in interior-point methods for structured problems, Research seminar of The Computer and Automation Research Institute , Hungarian Academy of Sciences, Budapest (Hungary), April 17 2008. Invited presentation.
  • J. Castro Interior point methods for LP, QP and convex problems: theory, implementation, and applications, First Intensive School on Mathematical Programming and Applications, CIEM Castro-Urdiales (Spain), January 30--31 2008. Invited presentation.
  • S. Giessing, A. Hundepool and J. Castro, Rounding methods for protecting EU-aggregates , Joint UNECE/Eurostat Work Session on Statistical Data Confidentiality, Manchester (United Kingdom), December 2007. Invited presentation.
  • J. Castro and A. Ouorou, An interior-point approach for convex routing problems in data telecommunication networks, Second Mathematical Programming Society International Conference on Continuous Optimization , McMaster University, Hamilton (Ontario, Canada), August 2007. Invited presentation.
  • J. Castro and J. Cuesta, Improving preconditioners in interior-point methods for optimization through quadratic regularizations, 2007 International Conference On Preconditioning Techniques For Large Sparse Matrix Problems In Scientific And Industrial Applications, Toulouse (France), July 2007.

Top