Advertisement

Numerische Mathematik

, Volume 12, Issue 1, pp 83–90 | Cite as

Scaling of matrices to achieve specified row and column sums

  • Albert W. Marshall
  • Ingram Olkin
Article

Abstract

IfA is ann ×n matrix with strictly positive elements, then according to a theorem ofSinkhorn, there exist diagonal matricesD1 andD2 with strictly positive diagonal elements such thatD1A D2 is doubly stochastic. This note offers an alternative proof of a generalization due toBrualdi, Parter andScheider, and independently toSinkhorn andKnopp, who show that A need not be strictly positive, but only fully indecomposable. In addition, we show that the same scaling is possible (withD1 =D2) whenA is strictly copositive, and also discuss related scaling for rectangular matrices. The proofs given show thatD1 andD2 can be obtained as the solution of an appropriate extremal problem.

The scaled matrixD1A D2 is of interest in connection with the problem of estimating the transition matrix of a Markov chain which is known to be doubly stochastic. The scaling may also be of interest as an aid in numerical computations.

Keywords

Markov Chain Numerical Computation Mathematical Method Diagonal Element Transition Matrix 
These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

Preview

Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.

References

  1. Bauer, F. L.: Optimally scaled matrices. Numer. Math.5, 73–87 (1963)Google Scholar
  2. Baumert, L. D.: Extreme copositive quadratic forms. Pacific J. Math.19, 197–204 (1966).Google Scholar
  3. ——: Extreme copositive quadratic forms, 11. Pacific J. Math.20, 1–20 (1967)Google Scholar
  4. Brualdi, R. A., S. V. Parter andH. Schneider: The diagonal equivalence of a nonnegative matrix to a stochastic matrix. J. Math. Anal. Appl.16, 31–50 (1966).Google Scholar
  5. Goldstein, A. A., andJ. F. Price: An effective algorithm for minimization. Numer. Math.10, 184–189 (1967).Google Scholar
  6. Hobby, C., andR. Pyke: Doubly stochastic operators obtained from positive operators. Pacific J. Math.15, 153–157 (1965).Google Scholar
  7. Householder, A. S.: The Theory of Matrices in Numerical Analysis. New York: Blaisdell Publ. Co. 1964.Google Scholar
  8. Marcus, M., andM. Newman: The permanent of a symmetric matrix. Notices Am. Math. Soc.8, 595, # 587–85 (1961).Google Scholar
  9. —— ——: Generalized functions of symmetric matrices. Proc. Am. Math. Soc.16, 826–830 (1965).Google Scholar
  10. Marshall, A. W., andI. Olkin: Multivariate Chebyshev inequalities. Ann. Math. Statist.31, 1001–1014 (1960).Google Scholar
  11. Maxfield, J., andH. Minc: A doubly stochastic matrix equivalent to a given matrix. Notices Am. Math. Soc.9, 309, # 592–81 (1962).Google Scholar
  12. Menon, M. V.: Reduction of a matrix with positive elements to a doubly stochastic matrix. Proc. Am. Math. Soc.18, 244–247 (1967).Google Scholar
  13. Ryser, H. J.: Combinatorial Mathematics. The Mathematical Assoc. of Amer. (9963).Google Scholar
  14. Sinkhorn, R.: On best doubly stochastic estimates. Notices Am. Math. Soc.9, 294, * 592–34 (1962).Google Scholar
  15. ——: A relationship between arbitrary positive matrices and doubly stochastic matrices. Ann. Math. Statist.35, 876–879 (1964).Google Scholar
  16. ——: A relationship between arbitrary positive matrices and stochastic matrices. Canad. J. Math.18, 303–306 (1966).Google Scholar
  17. ——: Diagonal equivalence to matrices with prescribed row and column sums. Am. Math. Monthly74, 402–405 (1967).Google Scholar
  18. ——, andP. Knopp: Concerning nonnegative matrices and doubly stochastic matrices. Pacific J. Math.21, 343–348 (1967).Google Scholar

Copyright information

© Springer-Verlag 1968

Authors and Affiliations

  • Albert W. Marshall
    • 1
  • Ingram Olkin
    • 2
  1. 1.Boeing Scientific Research LaboratoriesSeattle
  2. 2.Department of Statistics Stanford UniversityStanford

Personalised recommendations