Accelerating EM: An Empirical Study
نویسندگان
چکیده
Many applications require that we learn the pa rameters of a model from data. EM (E xpectation Maximization) is a method for learning the pa rameters of probabilistic models with missing or hidden data. There are instances in which this method is slow to converge. Therefore, sev eral accelerations have been proposed to improve the method. None of the proposed acceleration methods are theoretically dominant and experi mental comparisons are lacking. In this paper, we present the different proposed accelerations and compare them experimentally. From the re sults of the experiments, we argue that some acceleration of EM is always possible, but that which acceleration is superior depends on prop erties of the problem.
منابع مشابه
Structure-Activity Relationship of Imidazobenzodiazepines, an AM1 Semi-Empirical Quantum Mechanics Study
Conformations and electronic properties of a series of imidazobenzodiazepines are investigated by AM1 semi-empirical quantum mechanics method. It is shown that substitution of Cl in position 7 instead of 8, changes the geometry of the seven membered lactam ring; this may put the N5 nitrogen in a better positon to act as a hydrogen bond acceptor, and the phenyl ring in position 6 is probably...
متن کاملDispersion Engineering and Disorder in Photonic Crystals for Accelerator Applications
The possibility of achieving higher accelerating gradients at higher frequencies with the reduction of the effect of HOMs, compared to conventional accelerating structures, is increasing interest in the possible use of Photonic Crystals (PC) for accelerator applications. In this paper we analyze how the properties of the lattice of a PC resonator can be engineered to give a specific band struct...
متن کاملAn Application of the EM-algorithm to Approximate Empirical Distributions of Financial Indices with the Gaussian Mixtures
In this study I briefly illustrate application of the Gaussian mixtures to approximate empirical distributions of financial indices (DAX, Dow Jones, Nikkei, RTSI, S&P 500). The resulting distributions illustrate very high quality of approximation as evaluated by Kolmogorov-Smirnov test. This implies further study of application of the Gaussian mixtures to approximate empirical distributions of ...
متن کاملOn Aitken's Method and Other Approaches for Accelerating Convergence of the Em Algorithm
The EM algorithm of Dempster, Laird, and Rubin (1977) is a broadly applicable approach that has been widely applied to the iterative computation of maximum likelihood estimates in a variety of incomplete-data problems. A criticism that has been levelled at the EM algorithm is that its convergence can be quite slow. Unfortunately, methods to accelerate the EM algorithm do tend to sacriice the si...
متن کاملConvergence study of an accelerated ML-EM algorithm using bigger step size.
In SPECT/PET, the maximum-likelihood expectation-maximization (ML-EM) algorithm is getting more attention as the speed of computers increases. This is because it can incorporate various physical aspects into the reconstruction process leading to a more accurate reconstruction than other analytical methods such as filtered-backprojection algorithms. However, the convergence rate of the ML-EM alg...
متن کامل