Iterated Proportional Fitting Procedure and Infinite Products of Stochastic Matrices

DOI10.1007/978-3-319-92420-5_3MaRDI QIDQ5126589zbMATH OpenOpenAlexFDO

Authors Jean Brossard, Christophe Leuridan

Publication date 20 October 2020

Published in Séminaire de Probabilités XLIX (Search for Journal in Brave)

Full work available at URL https://arxiv.org/abs/1606.09126

relative entropy iterative proportional fitting contingency matrices distributions with given marginals infinite products of stochastic matrices

Mathematics Subject Classification ID

Contingency tables (62H17) Random matrices (probabilistic aspects) (60B20) Positive matrices and their generalizations; cones of matrices (15B48) Stochastic matrices (15B51)

Abstract: The iterative proportional fitting procedure, introduced in 1937 by Kruithof, aims to adjust the elements of an array to satisfy specified row and column sums. Given a rectangular non-negative matrix

X_{0}

and two positive marginals

a

and

b

, the algorithm generates a sequence of matrices

(X_{n})

starting at

X_{0}

, supposed to converge to a biproportional fitting, that is, to a matrix

Y

whose marginals are

a

and

b

and of the form

Y = D_{1} X_{0} D_{2}

, for some diagonal matrices

D_{1}

and

D_{2}

with positive diagonal entries. When a biproportional fitting does exist, it is unique and the sequence

(X_{n})

converges to it at an at least geometric rate. More generally, when there exists some matrix with marginal

a

and

b

and with support included in the support of

X_{0}

, the sequence

(X_{n})

converges to the unique matrix whose marginals are

a

and

b

and which can be written as a limit of matrices of the form

D_{1} X_{0} D_{2}

. In the opposite case, the sequence

(X_{n})

diverges but both subsequences

(X_{2 n})

and

(X_{2 n + 1})

converge. In the present paper, we use a new method to prove again these results and determine the two limit-points in the case of divergence. Our proof relies on a new convergence theorem for backward infinite products

c d o t s M_{2} M_{1}

of stochatic matrices

M_{n}

, with diagonal entries

M_{n} (i, i)

bounded away from

0

and with bounded ratios

M_{n} (j, i) / M_{n} (i, j)

. This theorem generalizes Lorenz' stabilization theorem. We also provide an alternative proof of Touric and Nedi'c's theorem on backward infinite products of doubly-stochatic matrices, with diagonal entries bounded away from

0

. In both situations, we improve slightly the conclusion, since we establish not only the convergence of the sequence

(M_{n} c d o t s M_{1})

, but also its finite variation.

Recommendations

Cites work

Cited in

(5)

This page was built for publication: Iterated Proportional Fitting Procedure and Infinite Products of Stochastic Matrices

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q5126589)