Anqi Dong , Tryphon T. Georgiou , Allen Tannenbaum
{"title":"符号不定先验的数据同化:Sinkhorn算法的推广","authors":"Anqi Dong , Tryphon T. Georgiou , Allen Tannenbaum","doi":"10.1016/j.automatica.2025.112283","DOIUrl":null,"url":null,"abstract":"<div><div>We deal with a hitherto unstudied problem that is motivated by the need to model promotion/inhibition contributions of relations, as in gene regulatory networks, to explain observed empirical distributions. There is no known pre-existing work focusing on this problem which, in essence, consists in calibrating a sign-indefinite data set representing prior information on the joint strength of promotion/inhibition, to restore consistency with specified observed marginals. We address the problem in the general setting of relations between <span><math><mi>n</mi></math></span> objects, and then specialize to binary relations to compare with a possible adhoc alternative following a Metropolis–Hastings rationale. Thus, our paper begins by addressing the general problem to find a multi-dimensional array (representing relative contribution of the <span><math><mi>n</mi></math></span> factors) to match specified marginals; the salient feature is that the sign (promotion/inhibition) of entries should match that of the prior. We develop an approach that is based on minimizing a suitable entropy functional subject to the constraints. Our approach follows verbatim Schrödinger’s rationale (Schrödinger, 1931, 1932) that underlies the celebrated Sinkhorn–Knopp algorithm in statistics used to reconcile contingency tables (of positive/frequency data sets) with marginal data (Knight, 2008). The resulting algorithm for our case of sign-indefinite prior generalizes the Sinkhorn–Knopp algorithm in that it amounts to iterative scaling of the entries of the array along different coordinate directions. The scaling is <em>multiplicative</em> but also, in contrast to Sinkhorn–Knopp, <em>inverse-multiplicative</em> depending on the sign of entries. When the entries of the prior are non-negative, our algorithm reduces to the classical Sinkhorn–Knopp algorithm which, in recent years, has become the computational workhorse of optimal transport (Peyré and Cuturi, 2019). Interestingly, the extension of the Sinkhorn–Knopp algorithm herein is perhaps the only example of a coordinate gradient ascent algorithm, other than the classical Sinkhorn, that has a non-trivial closed-form solution.</div></div>","PeriodicalId":55413,"journal":{"name":"Automatica","volume":"177 ","pages":"Article 112283"},"PeriodicalIF":5.9000,"publicationDate":"2025-04-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"Data Assimilation for Sign-indefinite Priors: A generalization of Sinkhorn’s algorithm\",\"authors\":\"Anqi Dong , Tryphon T. Georgiou , Allen Tannenbaum\",\"doi\":\"10.1016/j.automatica.2025.112283\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"<div><div>We deal with a hitherto unstudied problem that is motivated by the need to model promotion/inhibition contributions of relations, as in gene regulatory networks, to explain observed empirical distributions. There is no known pre-existing work focusing on this problem which, in essence, consists in calibrating a sign-indefinite data set representing prior information on the joint strength of promotion/inhibition, to restore consistency with specified observed marginals. We address the problem in the general setting of relations between <span><math><mi>n</mi></math></span> objects, and then specialize to binary relations to compare with a possible adhoc alternative following a Metropolis–Hastings rationale. Thus, our paper begins by addressing the general problem to find a multi-dimensional array (representing relative contribution of the <span><math><mi>n</mi></math></span> factors) to match specified marginals; the salient feature is that the sign (promotion/inhibition) of entries should match that of the prior. We develop an approach that is based on minimizing a suitable entropy functional subject to the constraints. Our approach follows verbatim Schrödinger’s rationale (Schrödinger, 1931, 1932) that underlies the celebrated Sinkhorn–Knopp algorithm in statistics used to reconcile contingency tables (of positive/frequency data sets) with marginal data (Knight, 2008). The resulting algorithm for our case of sign-indefinite prior generalizes the Sinkhorn–Knopp algorithm in that it amounts to iterative scaling of the entries of the array along different coordinate directions. The scaling is <em>multiplicative</em> but also, in contrast to Sinkhorn–Knopp, <em>inverse-multiplicative</em> depending on the sign of entries. When the entries of the prior are non-negative, our algorithm reduces to the classical Sinkhorn–Knopp algorithm which, in recent years, has become the computational workhorse of optimal transport (Peyré and Cuturi, 2019). Interestingly, the extension of the Sinkhorn–Knopp algorithm herein is perhaps the only example of a coordinate gradient ascent algorithm, other than the classical Sinkhorn, that has a non-trivial closed-form solution.</div></div>\",\"PeriodicalId\":55413,\"journal\":{\"name\":\"Automatica\",\"volume\":\"177 \",\"pages\":\"Article 112283\"},\"PeriodicalIF\":5.9000,\"publicationDate\":\"2025-04-08\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"Automatica\",\"FirstCategoryId\":\"94\",\"ListUrlMain\":\"https://www.sciencedirect.com/science/article/pii/S000510982500175X\",\"RegionNum\":2,\"RegionCategory\":\"计算机科学\",\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"Q1\",\"JCRName\":\"AUTOMATION & CONTROL SYSTEMS\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"Automatica","FirstCategoryId":"94","ListUrlMain":"https://www.sciencedirect.com/science/article/pii/S000510982500175X","RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q1","JCRName":"AUTOMATION & CONTROL SYSTEMS","Score":null,"Total":0}
Data Assimilation for Sign-indefinite Priors: A generalization of Sinkhorn’s algorithm
We deal with a hitherto unstudied problem that is motivated by the need to model promotion/inhibition contributions of relations, as in gene regulatory networks, to explain observed empirical distributions. There is no known pre-existing work focusing on this problem which, in essence, consists in calibrating a sign-indefinite data set representing prior information on the joint strength of promotion/inhibition, to restore consistency with specified observed marginals. We address the problem in the general setting of relations between objects, and then specialize to binary relations to compare with a possible adhoc alternative following a Metropolis–Hastings rationale. Thus, our paper begins by addressing the general problem to find a multi-dimensional array (representing relative contribution of the factors) to match specified marginals; the salient feature is that the sign (promotion/inhibition) of entries should match that of the prior. We develop an approach that is based on minimizing a suitable entropy functional subject to the constraints. Our approach follows verbatim Schrödinger’s rationale (Schrödinger, 1931, 1932) that underlies the celebrated Sinkhorn–Knopp algorithm in statistics used to reconcile contingency tables (of positive/frequency data sets) with marginal data (Knight, 2008). The resulting algorithm for our case of sign-indefinite prior generalizes the Sinkhorn–Knopp algorithm in that it amounts to iterative scaling of the entries of the array along different coordinate directions. The scaling is multiplicative but also, in contrast to Sinkhorn–Knopp, inverse-multiplicative depending on the sign of entries. When the entries of the prior are non-negative, our algorithm reduces to the classical Sinkhorn–Knopp algorithm which, in recent years, has become the computational workhorse of optimal transport (Peyré and Cuturi, 2019). Interestingly, the extension of the Sinkhorn–Knopp algorithm herein is perhaps the only example of a coordinate gradient ascent algorithm, other than the classical Sinkhorn, that has a non-trivial closed-form solution.
期刊介绍:
Automatica is a leading archival publication in the field of systems and control. The field encompasses today a broad set of areas and topics, and is thriving not only within itself but also in terms of its impact on other fields, such as communications, computers, biology, energy and economics. Since its inception in 1963, Automatica has kept abreast with the evolution of the field over the years, and has emerged as a leading publication driving the trends in the field.
After being founded in 1963, Automatica became a journal of the International Federation of Automatic Control (IFAC) in 1969. It features a characteristic blend of theoretical and applied papers of archival, lasting value, reporting cutting edge research results by authors across the globe. It features articles in distinct categories, including regular, brief and survey papers, technical communiqués, correspondence items, as well as reviews on published books of interest to the readership. It occasionally publishes special issues on emerging new topics or established mature topics of interest to a broad audience.
Automatica solicits original high-quality contributions in all the categories listed above, and in all areas of systems and control interpreted in a broad sense and evolving constantly. They may be submitted directly to a subject editor or to the Editor-in-Chief if not sure about the subject area. Editorial procedures in place assure careful, fair, and prompt handling of all submitted articles. Accepted papers appear in the journal in the shortest time feasible given production time constraints.