Approximate Expectation Maximization
نویسندگان
چکیده
We discuss the integration of the expectation-maximization (EM) algorithm for maximum likelihood learning of Bayesian networks with belief propagation algorithms for approximate inference. Specifically we propose to combine the outer-loop step of convergent belief propagation algorithms with the M-step of the EM algorithm. This then yields an approximate EM algorithm that is essentially still double loop, with the important advantage of an inner loop that is guaranteed to converge. Simulations illustrate the merits of such an approach.
منابع مشابه
Sigma-Point Filtering and Smoothing Based Parameter Estimation in Nonlinear Dynamic Systems
We consider approximate maximum likelihood parameter estimation in nonlinear state-space models. We discuss both direct optimization of the likelihood and expectation– maximization (EM). For EM, we also give closed-form expressions for the maximization step in a class of models that are linear in parameters and have additive noise. To obtain approximations to the filtering and smoothing distrib...
متن کاملThe Development of Maximum Likelihood Estimation Approaches for Adaptive Estimation of Free Speed and Critical Density in Vehicle Freeways
The performance of many traffic control strategies depends on how much the traffic flow models have been accurately calibrated. One of the most applicable traffic flow model in traffic control and management is LWR or METANET model. Practically, key parameters in LWR model, including free flow speed and critical density, are parameterized using flow and speed measurements gathered by inductive ...
متن کاملEstimating Bohm’s quantum force using Bayesian statistics
In this paper we develop an approximate methodology for estimating the multidimensional quantum density associated with a statistical bundle of de Broglie-Bohm trajectories. The quantum density is constructed as a discrete sum of nonequivalent Gaussian components. We incorporate the ideas of Bayesian statistical analysis and an expectation-maximization procedure to compute an approximate quantu...
متن کاملApproximate minimization algorithms for the 0/1 Knapsack and Subset-Sum Problem
The well-studied 0=1 Knapsack and Subset-Sum Problem are maximization problems that have an equivalent minimization version. While exact algorithms for one of these two versions also yield an exact solution for the other version, this does not apply to -approximate algorithms. We present several -approximate Greedy Algorithms for the minimization version of the 0=1 Knapsack and the Subset-Sum P...
متن کامل1 6 A ug 2 00 6 Functions for relative maximization
We introduce functions for relative maximization in a general context: the beta and alpha applications. After a systematic study concerning regularities, we investigate how to approximate certain values of these functions using periodic orbits. We establish yet that the differential of an alpha application dictates the asymptotic behavior of the optimal trajectories.
متن کامل