Abstract
Given an n × m nonnegative matrix A = (a ij ) and positive integral vectors \(r \in \mathfrak{R}^{n}\) and \(c \in \mathfrak{R}^{m}\) having a common one-norm h, the (r,c)-scaling problem is to obtain positive diagonal matrices X and Y, if they exist, such that XAY has row and column sums equal to r and c, respectively. The entropy minimization problem corresponding to A is to find an n × m matrix z = (z ij ) having the same zero pattern as A, the sum of whose entries is a given number h, its row and column sums are within given integral vectors of lower and upper bounds, and such that the entropy function consisting of the sum of the terms z ij ln (z ij /a ij ) is minimized. When the lower and upper bounds coincide, matrix scaling and entropy minimization are closely related. In this paper we present several complexity bounds for the \(\epsilon\)-approximate (r,c)-scaling problem, polynomial in n,m,h, \(1/\epsilon\), and ln \({V} \over {v}\), where V and v are the largest and the smallest positive entries of A, respectively. These bounds, although not polynomial in \(\ln (1/\epsilon)\), not only provide alternative complexities for the polynomial time algorithms, but could result in better overall complexities. In particular, our theoretical analysis supports the practicality of the well-known RAS algorithm. In our analysis we obtain bounds on the norm of scaling vectors which will be used in deriving not only some of the above complexities, but also a complexity for square nonnegative matrices having positive permanent. In particular, our results extend, nontrivially, many bounds for the doubly stochastic scaling of square nonnegative matrices previously given by Kalantari and Khachiyan to the case of general (r,c)-scaling. Finally, we study a more general entropy minimization problem where row and column sums are constrained to lie in prescribed intervals, and the sum of all entries is also prescribed. Balinski and Demange described an RAS type algorithm for its continuous version, but did not analyze its complexity. We show that this algorithm produces an \(\epsilon\)-approximate solution within complexity polynomial in n, m, h, \({V} \over {v}\) and \(1/\epsilon\).
Similar content being viewed by others
References
Bacharach M. Biproportional matrices and input–output change. Cambridge (1970)
Bachem A., Korte B (1979) On the RAS algorithm. Computing 23, 189–198
Balinski M.L. Demange G. (1989) An axiomatic approach to proportionality between matrices. Math. Oper. Res. 14, 700–719
Balinski M.L., Demange G. (1989) Algorithms for proportional matrices in reals and integers. Math. Program. (Ser. B) 45, 193–210
Bertsekas D.P., Polymenakos L.C., Tseng P. Epsilon-Relaxation and Auction Methods for Separable Convex Cost Network Flow Problems in Network Optimization. In: Pardalos P.M., Hearn D.W., Hager W.W. (eds) Proceedings of International Conference on Network Optimization. Springer, Berlin Heidelberg New York, 1998, pp 103–126 (1989)
Birkhoff G. (1957) Extensions of Jentzsch’s theorem. Trans. Am. Math. Soc. 85, 219–227
Bregman L.M. (1967) Proof of convergence of Sheleikhovskii’s method for a problem with transportation constraints. USSR Comput. Math. Math. Phys. 7, 191–204
Brualdi R.A., Prater S.V., Schneider H. (1966) The diagonal equivalence of a nonnegative matrix to a stochastic matrix. J. Math. Ann. Appl. 16, 31–50
Brualdi R.A. (1968) Convex sets of nonnegative matrices. Can. J. Math. 20, 144–157
Franklin J., Lorenz J. (1989) On the scaling of multidimensional matrices. Linear Algebra Appl. 114/115: 717–735
Hochbaum D.S. (1993) Polynomial and strongly polynomial algorithms for convex network optimization. In: Du D.Z., Pardalos P. (eds) Network Optimization Problems. World Scientific, Singapore
Kalantari, B.: Canonical problems for quadratic programming and projective methods for their solution. In: Lagarias, J.C., Todd, M. (eds.) Contemporary Mathematics. vol. 114, pp. 243–263. Proceedings of AMS conference “Mathematical problems arising from linear programming”, 1988 (1990)
Kalantari B. (1995) A simple polynomial time algorithm for a convex hull problem equivalent to linear programming. In: Colbourn C.J., Mahmoodian E.S. (eds). Combinatorics Advances. Kluwer, Dordrecht, pp. 207–216
Kalantari B. (1996) A theorem of the alternative for multihomogeneous functions and its relationship to diagonal scaling of matrices. Linear Algebra Appl. 236, 1–24
Kalantari B. (2003) Semidefinite Programming and Matrix Scaling over the Semidefinite Cone. Linear Algebra Appl. 375, 221–243
Kalantari, B.: Scaling dualities and self-concordant homogeneous programming in finite dimensional spaces. Technical Report LCSR-TR-359, Department of Computer Science, Rutgers University, New Brunswick (1999)
Kalantari B., Khachiyan L. (1993) On the rate of convergence of deterministic and randomized RAS matrix scaling method. Oper. Res. Lett. 14, 237–244
Kalantari B., Khachiyan L. (1996) On the complexity of nonnegative-matrix scaling. Linear Algebra Appl. 240, 87–103
Kalantari B., Khachiyan L., Shokoufandeh A. (1997) On the complexity of matrix balancing. SIAM J. Matrix Anal. Appl. 16, 450–463
Kalantari B., Emamy-K M.R. (1997) On linear programming and matrix scaling over the algebraic numbers. Linear Algebra Appl. 262, 283–306
Kalantari, B., Lari, I., Ricca, F., Simeone, B.: On the complexity of general matrix scaling and entropy minimization via the RAS algorithm. Technical Report, n.24, Department of Statistics and Applied probability, La Sapienza University, Rome (2002)
Karmarkar N. (1984) A new polynomial time algorithm for linear programming. Combinatorica 4, 373–395
Karzanov A.V. McCormick S.T. (1997) Polynomial methods for separable convex optimization in unimodular linear spaces with applications. SIAM J. Comput. 26, 1245–1275
Khachiyan L., Kalantari B. (1992) Diagonal matrix scaling and linear programming. SIAM J. Optim. 4, 668–672
Linial N., Samorodnitsky A., Wigderson A. (2000) A deterministic strongly polynomial algorithm for matrix scaling and approximate permanents. Combinatorica 20, 545–568
Lovász L., Plummer M.D. (1986) Matching Theory. Annals of Discrete Mathematics, vol 121. North-Holland, Amsterdam
Marshall A.W., Olkin I. (1968) Scaling of matrices to achieve specified row and column sums. Numer. Math. 12, 83–90
Maxfield J., Minc H. (1962) A doubly stochastic matrix equivalent to a given matrix. Notices Am. Math. Soc. 9, 309
Menon M.V. (1967) Reduction of a matrix with positive elements to a doubly stochastic matrix. Proc. Am. Math. Soc. 18, 244–247
Minc H. (1988) Nonnegative Matrices. Wiley, New York
Nemirovski A., Rothblum U.G. (1999) On complexity of matrix scaling. Linear Algebra Appl. 302/303: 435–460
Nesterov Y., Nemirovskii A.S. (1994) Interior-Point Polynomial Algorithms in Convex Programming. SIAM, Philadelphia
Potra F., Ye Y. (1993) A quadratically convergent polynomial algorithm for solving entropy optimization problems. SIAM J. Optm. 4, 843–860
Rothblum U.G., Schneider H. (1989) Scaling of matrices which have prescribed row sums and column sums via optimization. Linear Algebra Appl. 114/115: 737–764
Rothblum U.G. (1989) Generalized scaling satisfying linear equations. Linear Algebra Appl. 114/115: 765–783
Schneider M.H., Zenios S.A. (1989) A comparative study of algorithms for matrix balancing. Oper. Res. 38, 439–455
Schneider M.H. (1989) Matrix scaling, entropy minimization, and conjugate duality. Linear Algebra Appl. 114/115: 785–813
Sinkhorn R. (1964) A relationship between arbitrary positive matrices and doubly stochastic matrices. Ann. Math. Statist. 35, 876–879
Author information
Authors and Affiliations
Corresponding author
Rights and permissions
About this article
Cite this article
Kalantari, B., Lari, I., Ricca, F. et al. On the complexity of general matrix scaling and entropy minimization via the RAS algorithm. Math. Program. 112, 371–401 (2008). https://doi.org/10.1007/s10107-006-0021-4
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10107-006-0021-4