EM Algorithms: A New Perspective∗
نویسنده
چکیده
The EM algorithm is not a single algorithm, but a template for the construction of iterative algorithms. While it is always presented in stochastic language, relying on conditional expectations to obtain a method for estimating parameters in statistics, the essence of the EM algorithm is not stochastic. The conventional formulation of the EM algorithm given in many texts and papers on the subject is inadequate. A new formulation is given here based on the notion of acceptable data.
منابع مشابه
EM Algorithms from a Non-stochastic Perspective
The EM algorithm is not a single algorithm, but a template for the construction of iterative algorithms. While it is always presented in stochastic language, relying on conditional expectations to obtain a method for estimating parameters in statistics, the essence of the EM algorithm is not stochastic. The conventional formulation of the EM algorithm given in many texts and papers on the subje...
متن کاملBreast Cancer Diagnosis from Perspective of Class Imbalance
Introduction: Breast cancer is the second cause of mortality among women. Early detection is the only rescue to reduce the risk of breast cancer mortality. Traditional methods cannot effectively diagnose tumor since they are based on the assumption of well-balanced dataset.. However, a hybrid method can help to alleviate the two-class imbalance problem existing in the ...
متن کاملEM algorithms without missing data.
Most problems in computational statistics involve optimization of an objective function such as a loglikelihood, a sum of squares, or a log posterior function. The EM algorithm is one of the most effective algorithms for maximization because it iteratively transfers maximization from a complex function to a simple, surrogate function. This theoretical perspective clarifies the operation of the ...
متن کاملModelling State in Mind and Machine
This paper discusses and illustrates the application of Empirical Modelling (EM), as developed by the author and his collaborators, to computer programming. The discussion centres on the perspective EM gives on procedural and experiential vs. declarative and logical approaches to knowledge representation. The illustration focuses on demonstrating the potential for pedagogical use of EM in teach...
متن کاملNew global optimization algorithms for model-based clustering
The Expectation-Maximization (EM) algorithm is a very popular optimization tool in model-based clustering problems. However, while the algorithm is convenient to implement and numerically very stable, it only produces solutions that are locally optimal. Thus, EM may not achieve the globally optimal solution to clustering problems, which can have a large number of local optima. This paper introd...
متن کامل