Expectation Maximization as Message Passing - Part I: Principles and Gaussian Messages Export

@article{citeulike:5951177, abstract = {It is shown how expectation maximization (EM) may be viewed as a message passing algorithm in factor graphs. In particular, a general EM message computation rule is identified. As a factor graph tool, EM may be used to break cycles in a factor graph, and tractable messages may in some cases be obtained where the sum-product messages are unwieldy. As an exemplary application, the paper considers linear Gaussian state space models. Unknown coefficients in such models give rise to multipliers in the corresponding factor graph. A main attraction of EM in such cases is that it results in purely Gaussian message passing algorithms. These Gaussian EM messages are tabulated for several (scalar, vector, matrix) multipliers that frequently appear in applications.}, archivePrefix = {arXiv}, author = {Dauwels, Justin and Eckford, Andrew and Korl, Sascha and Loeliger, Hans-Andrea}, citeulike-article-id = {5951177}, citeulike-linkout-0 = {http://arxiv.org/abs/0910.2832}, citeulike-linkout-1 = {http://arxiv.org/pdf/0910.2832}, day = {15}, eprint = {0910.2832}, keywords = {distributedalgorithms, messagepassingalgorithms, signalprocessing, statistics}, month = {Oct}, posted-at = {2009-10-16 14:16:03}, priority = {2}, title = {Expectation Maximization as Message Passing - Part I: Principles and Gaussian Messages}, url = {http://arxiv.org/abs/0910.2832}, year = {2009} }

TY - JOUR ID - citeulike:5951177 L3 - citeulike-article-id:5951177 N2 - It is shown how expectation maximization (EM) may be viewed as a message passing algorithm in factor graphs. In particular, a general EM message computation rule is identified. As a factor graph tool, EM may be used to break cycles in a factor graph, and tractable messages may in some cases be obtained where the sum-product messages are unwieldy. As an exemplary application, the paper considers linear Gaussian state space models. Unknown coefficients in such models give rise to multipliers in the corresponding factor graph. A main attraction of EM in such cases is that it results in purely Gaussian message passing algorithms. These Gaussian EM messages are tabulated for several (scalar, vector, matrix) multipliers that frequently appear in applications. TI - Expectation Maximization as Message Passing - Part I: Principles and Gaussian Messages KW - distributedalgorithms KW - messagepassingalgorithms KW - signalprocessing KW - statistics AU - Dauwels, Justin AU - Eckford, Andrew AU - Korl, Sascha AU - Loeliger, Hans-Andrea PY - 2009/Oct/15/ UR - http://arxiv.org/abs/0910.2832 ER -