The Experts below are selected from a list of 63675 Experts worldwide ranked by ideXlab platform
Hans-andrea Loeliger - One of the best experts on this subject based on the ideXlab platform.
-
ISIT - Particle Methods as Message Passing
2006 IEEE International Symposium on Information Theory, 2006Co-Authors: Justin Dauwels, Sascha Korl, Hans-andrea LoeligerAbstract:It is shown how particle methods can be viewed as Message Passing on factor graphs. In this setting, particle methods can readily be combined with other Message-Passing techniques such as the sum-product and max-product algorithm, expectation maximization, iterative conditional modes, steepest descent, Kalman filters, etc. Generic Message computation rules for particle-based representations of sum-product Messages are formulated. Various existing particle methods are described as instances of those generic rules, i.e., Gibbs sampling, importance sampling, Markov-chain Monte Carlo methods (MCMC), particle filtering, and simulated annealing.
-
Expectation maximization as Message Passing
Proceedings. International Symposium on Information Theory 2005. ISIT 2005., 2005Co-Authors: Justin Dauwels, S. Korl, Hans-andrea LoeligerAbstract:Based on prior work by Eckford, it is shown how expectation maximization (EM) may be viewed, and used, as a Message Passing algorithm in factor graphs.
Vladimir Kolmogorov - One of the best experts on this subject based on the ideXlab platform.
-
A New Look at Reweighted Message Passing
IEEE transactions on pattern analysis and machine intelligence, 2015Co-Authors: Vladimir KolmogorovAbstract:We propose a new family of Message Passing techniques for MAP estimation in graphical models which we call Sequential Reweighted Message Passing (SRMP). Special cases include well-known techniques such as Min-Sum Diffusion (MSD) and a faster Sequential Tree-Reweighted Message Passing (TRW-S). Importantly, our derivation is simpler than the original derivation of TRW-S, and does not involve a decomposition into trees. This allows easy generalizations. The new family of algorithms can be viewed as a generalization of TRW-S from pairwise to higher-order graphical models. We test SRMP on several real-world problems with promising results.
-
Reweighted Message Passing revisited.
arXiv: Artificial Intelligence, 2013Co-Authors: Vladimir KolmogorovAbstract:We propose a new family of Message Passing techniques for MAP estimation in graphical models which we call {\em Sequential Reweighted Message Passing} (SRMP). Special cases include well-known techniques such as {\em Min-Sum Diffusion} (MSD) and a faster {\em Sequential Tree-Reweighted Message Passing} (TRW-S). Importantly, our derivation is simpler than the original derivation of TRW-S, and does not involve a decomposition into trees. This allows easy generalizations. We present such a generalization for the case of higher-order graphical models, and test it on several real-world problems with promising results.
Qifeng Huang - One of the best experts on this subject based on the ideXlab platform.
-
ws based workflow description language for Message Passing
Cluster Computing and the Grid, 2005Co-Authors: Yan Huang, Qifeng HuangAbstract:This paper presents MPFL, an extension to SWFL that provides a Message Passing flow language for describing parallel scientific applications with Message Passing composed from Web services. The background to developing the language is discussed. Then the paper focuses on four main Message Passing issues: communicators, point-to-point communications, collective communications, and application topologies. The semantics of the language and its model for representing Message Passing are also described.
-
CCGRID - WS-based workflow description language for Message Passing
CCGrid 2005. IEEE International Symposium on Cluster Computing and the Grid 2005., 2005Co-Authors: Yan Huang, Qifeng HuangAbstract:This paper presents MPFL, an extension to SWFL that provides a Message Passing flow language for describing parallel scientific applications with Message Passing composed from Web services. The background to developing the language is discussed. Then the paper focuses on four main Message Passing issues: communicators, point-to-point communications, collective communications, and application topologies. The semantics of the language and its model for representing Message Passing are also described.
Justin Dauwels - One of the best experts on this subject based on the ideXlab platform.
-
ISIT - Particle Methods as Message Passing
2006 IEEE International Symposium on Information Theory, 2006Co-Authors: Justin Dauwels, Sascha Korl, Hans-andrea LoeligerAbstract:It is shown how particle methods can be viewed as Message Passing on factor graphs. In this setting, particle methods can readily be combined with other Message-Passing techniques such as the sum-product and max-product algorithm, expectation maximization, iterative conditional modes, steepest descent, Kalman filters, etc. Generic Message computation rules for particle-based representations of sum-product Messages are formulated. Various existing particle methods are described as instances of those generic rules, i.e., Gibbs sampling, importance sampling, Markov-chain Monte Carlo methods (MCMC), particle filtering, and simulated annealing.
-
Expectation maximization as Message Passing
Proceedings. International Symposium on Information Theory 2005. ISIT 2005., 2005Co-Authors: Justin Dauwels, S. Korl, Hans-andrea LoeligerAbstract:Based on prior work by Eckford, it is shown how expectation maximization (EM) may be viewed, and used, as a Message Passing algorithm in factor graphs.
Joseph A. O'sullivan - One of the best experts on this subject based on the ideXlab platform.
-
Message Passing expectation-maximization algorithms
IEEE SP 13th Workshop on Statistical Signal Processing 2005, 2005Co-Authors: Joseph A. O'sullivanAbstract:Message Passing algorithms have had dramatic impacts on important problems in signal processing, learning theory, communication theory, and information theory through their computational efficiency. Expectation-maximization algorithms have had dramatic impacts on problems in estimation and detection theory, but their computational efficiency often limits their applicability. Given a bipartite graphical model for the data, if a set of hidden independent random variables can be associated with the edges, then a resulting expectation-maximization algorithm is Message Passing on this graph. The algorithms are computationally efficient in the same sense as other Message Passing algorithms. One example of such algorithms is the standard expectation-maximization algorithm for emission tomography. Another example for a signal in Gaussian noise yields a statistical interpretation to efficient algorithms for sparse linear inverse problems
-
Message Passing expectation-maximization algorithms
IEEE SP 13th Workshop on Statistical Signal Processing 2005, 2005Co-Authors: Joseph A. O'sullivanAbstract:Message Passing algorithms have had dramatic impacts on important problems in signal processing, learning theory, communication theory, and information theory through their computational efficiency. Expectation-maximization algorithms have had dramatic impacts on problems in estimation and detection theory, but their computational efficiency often limits their applicability. Given a bipartite graphical model for the data, if a set of hidden independent random variables can be associated with the edges, then a resulting expectation-maximization algorithm is Message Passing on this graph. The algorithms are computationally efficient in the same sense as other Message Passing algorithms. One example of such algorithms is the standard expectation-maximization algorithm for emission tomography. Another example for a signal in Gaussian noise yields a statistical interpretation to efficient algorithms for sparse linear inverse problems
-
Ordered subsets Message-Passing
IEEE International Symposium on Information Theory 2003. Proceedings., 2003Co-Authors: Joseph A. O'sullivan, N. SinglaAbstract:A Message-Passing algorithm is proposed for decoding on graphs having short cycles. The algo- rithm, termed the "ordered subsets Message-Passing" (OSMP) algorithm, performs Message-Passing on a graph in which the measured data is partitioned into subsets. The OSMP algorithm is applied for joint equalization and decoding for two-dimensional (2D) intersymbol interference (ISI) channels. Simulation results show that the OSMP algorithm outperforms its unordered counterpart i.e. when the measured data are not partitioned into subsets. Concentration results proved by KavEid et. al. in (2) for a one- dimensional IS1 channel also hold for the OSMP al- gorithm for a 2D IS1 channel.
