Fundamental Linear Algebra Problem of Gaussian Inference

Underlying many Bayesian inference techniques that seek to approximate the posterior as a Gaussian distribution is a fundamental linear algebra problem that must be solved for both the mean and key entries of the covariance. Even when the true posterior is not Gaussian (e.g., in the case of nonlinear measurement functions) we can use variational schemes that repeatedly solve this linear algebra problem at each iteration. In most cases, the question is not whether a solution to this problem exists, but rather how we can exploit problem-specific structure to find it efficiently. Our contribution is to clearly state the Fundamental Linear Algebra Problem of Gaussian Inference (FLAPOGI) and to provide a novel presentation (using Kronecker algebra) of the not-so-well-known result of Takahashi et al. (1973) that makes it possible to solve for key entries of the covariance matrix. We first provide a global solution and then a local version that can be implemented using local message passing amongst a collection of agents calculating in parallel. Contrary to belief propagation, our local scheme is guaranteed to converge in both the mean and desired covariance quantities to the global solution even when the underlying factor graph is loopy; in the case of synchronous updates, we provide a bound on the number of iterations required for convergence. Compared to belief propagation, this guaranteed convergence comes at the cost of additional storage, calculations, and communication links in the case of loops; however, we show how these can be automatically constructed on the fly using only local information.

[1]  Panganamala Ramana Kumar,et al.  Extended message passing algorithm for inference in loopy Gaussian graphical models , 2004, Ad Hoc Networks.

[2]  Timothy D. Barfoot,et al.  Multivariate Gaussian Variational Inference by Natural Gradient Descent , 2020, ArXiv.

[3]  Stefan B. Williams,et al.  Efficient View-Based SLAM Using Visual Loop Closures , 2008, IEEE Transactions on Robotics.

[4]  Frank Dellaert,et al.  iSAM2: Incremental smoothing and mapping with fluid relinearization and incremental variable reordering , 2011, 2011 IEEE International Conference on Robotics and Automation.

[5]  Soummya Kar,et al.  Convergence Analysis of Distributed Inference with Vector-Valued Gaussian Belief Propagation , 2016, J. Mach. Learn. Res..

[6]  Paul H. Siegel,et al.  Gaussian belief propagation solver for systems of linear equations , 2008, 2008 IEEE International Symposium on Information Theory.

[7]  Hugh Durrant-Whyte,et al.  Simultaneous Localisation and Mapping ( SLAM ) : Part I The Essential Algorithms , 2006 .

[8]  J. Magnus,et al.  The Commutation Matrix: Some Properties and Applications , 1979 .

[9]  Calcul partiel de l ’ inverse d ’ une matrice creuse de grande taille-application en astrophysique Partial computation of the inverse of a large sparse matrix-application to astrophysics , 2011 .

[10]  Judea Pearl,et al.  Probabilistic reasoning in intelligent systems , 1988 .

[11]  Timothy D. Barfoot,et al.  Variational Inference as Iterative Projection in a Bayesian Hilbert Space 2 Bayesian Hilbert Spaces , 2020 .

[12]  William T. Freeman,et al.  Correctness of Belief Propagation in Gaussian Graphical Models of Arbitrary Topology , 1999, Neural Computation.

[13]  Andrew J. Davison,et al.  FutureMapping 2: Gaussian Belief Propagation for Spatial AI , 2019, ArXiv.

[14]  Dirk Slock,et al.  Convergence Analysis of Sparse Bayesian Learning Under Approximate Inference Techniques , 2019, 2019 53rd Asilomar Conference on Signals, Systems, and Computers.

[15]  J. Magnus,et al.  Matrix Differential Calculus with Applications in Statistics and Econometrics , 1991 .

[16]  David Yoon,et al.  Exactly sparse Gaussian variational inference with application to derivative-free batch nonlinear state estimation , 2019, ArXiv.

[17]  W. F. Tinney,et al.  On computing certain elements of the inverse of a sparse matrix , 1975, Commun. ACM.

[18]  Fabio Verbosio High performance selected inversion methods for sparse matrices , 2019 .

[19]  Jan R. Magnus,et al.  The Elimination Matrix: Some Lemmas and Applications , 1980, SIAM J. Algebraic Discret. Methods.

[20]  Nir Friedman,et al.  Probabilistic Graphical Models - Principles and Techniques , 2009 .

[21]  Frank Dellaert,et al.  iSAM: Incremental Smoothing and Mapping , 2008, IEEE Transactions on Robotics.

[22]  Michael I. Jordan,et al.  An Introduction to Variational Methods for Graphical Models , 1999, Machine Learning.

[23]  Danny Bickson,et al.  Gaussian Belief Propagation: Theory and Aplication , 2008, 0811.2518.

[24]  TIMOTHY A. DAVISyTechnical,et al.  Computing the Sparse Inverse Subset: an Inverse Multifrontal Approach , 1995 .

[25]  Andrew W. Fitzgibbon,et al.  Bundle Adjustment - A Modern Synthesis , 1999, Workshop on Vision Algorithms.

[26]  David Bolin,et al.  Efficient Covariance Approximations for Large Sparse Precision Matrices , 2017, Journal of Computational and Graphical Statistics.

[27]  M. Yannakakis Computing the Minimum Fill-in is NP^Complete , 1981 .

[28]  Shun-ichi Amari,et al.  Natural Gradient Works Efficiently in Learning , 1998, Neural Computation.

[29]  พงศ์ศักดิ์ บินสมประสงค์,et al.  FORMATION OF A SPARSE BUS IMPEDANCE MATRIX AND ITS APPLICATION TO SHORT CIRCUIT STUDY , 1980 .

[30]  Huaiyu Zhu On Information and Sufficiency , 1997 .

[31]  Frank Dellaert,et al.  Incremental smoothing and mapping , 2008 .

[32]  Mathias Jacquelin,et al.  A Left-Looking Selected Inversion Algorithm and Task Parallelism on Shared Memory Systems , 2016, HPC Asia.

[33]  F. Alvarado,et al.  Parallel Inversion of Sparse Matrices. , 1986, IEEE Transactions on Power Systems.

[34]  Tianju Sui,et al.  Convergence and Accuracy Analysis for a Distributed Static State Estimator Based on Gaussian Belief Propagation , 2020, IEEE Transactions on Automatic Control.

[35]  F. Opitz Information geometry and its applications , 2012, 2012 9th European Radar Conference.

[36]  Amir Globerson,et al.  Convergent message passing algorithms - a unifying view , 2009, UAI.

[37]  Frank Dellaert,et al.  Covariance recovery from a square root information matrix for data association , 2009, Robotics Auton. Syst..

[38]  Tamir Hazan,et al.  Convergent Message-Passing Algorithms for Inference over General Graphs with Convex Free Energies , 2008, UAI.

[39]  Lexing Ying,et al.  SelInv---An Algorithm for Selected Inversion of a Sparse Symmetric Matrix , 2011, TOMS.

[40]  Manfred Opper,et al.  The Variational Gaussian Approximation Revisited , 2009, Neural Computation.