О выборе параметров ЕМ-алгоритма для разделения смеси распределений
Journal Title: Математичне моделювання - Year 2017, Vol 1, Issue 1
Abstract
ON THE CHOICE OF THE PARAMETERS OF THE EM ALGORITHM FOR THE SEPARATION OF THE MIXTURE OF DISTRIBUTIONS. Shumeiko A.A, Iskandarova-Mala A. Abstract The paper considers one of the types of non-hierarchical clustering - the EM-algorithm. This methd is traditionally the most used tool for separating a mixture of distributions with a known number of components. Determining the number of components (clusters) is a non-trivial problem. As a rule, this parameter is chosen by the researcher from some a priori assumptions. Nevertheless, the problem of the automatic choice of this parameter was investigated in many papers. Existing approaches, as a rule, differ in the complexity of implementation, intuitive approaches or approximate methods for solving complex analytical problems, it is often not clear how they are generally related to this task. The EM-algorithm for separating the mixture of normal distributions is in fact a reconstruction of the existing histogram by a linear combination of Gaussian functions, so, naturally, having the value of the specified error of reconstruction, describe the available histogram by a depleted histogram with free nodes, which is best (ie, with a minimum number nodes). The resulting number of nodes is an estimate of the number of components of the mixture, which is reconstructed by a linear combination of Gaussian functions with a given error. Moreover, the values of free nodes make it possible to obtain a starting estimate of the remaining parameters of the mixture. References [1]. McLachlan G.J. The EM algorithm and extensions / G. J. McLachlan, T. Krishnan.-New York: John Wiley & Sons, Inc., 1997 .– 288 p. [2]. Korolev V.Yu. EM-algorithm, its modifications and their application to the problem of separation of mixtures of probability distributions. / V.Yu.Korolev; Theoretical review .– M: IPI RAS, 2007 .– 94 p. [3]. Shumeiko A.A. Intelligent Data Analysis (Introduction to Data Mining) / A.A. Shumeiko, S.L.Sotnik .– Dnepropetrovsk: Belaya EA, 2012.– 212 pp. [4]. Optimal Histograms with Quality Guarantees / H.Jagadish, N.Koudas, S.Multhukrishan та ін. .– Access mode: http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.36.769&rep=rep1&type=ps [5]. Vetrov, D.P. Automatic determination of the number of components in the EM-algorithm for recovering a mixture of normal distributions / D.P.Vetrov, D.A.Kropotov, A.A.Asokin // Zh. Vychisl. Math. and Math. fiz. .– 2010 .– No. 4 (50).–P.770–783. [6]. Ligun AA Asymptotic methods for reconstructing curves / AA Ligun, AA Shumeiko .– K .: Naukova dumka, 1997 .– 358 pp.
Authors and Affiliations
А. А. Шумейко, А. О. Искандарова-Мала
Keywords
Related Articles
- EP ID EP277093
- DOI -
- Views 69
- Downloads 0
Математическая модель тепловой работы ошлакованной футеровки стен конвертера при взаимодействии ее с расплавом в ходе кислородной продувки
MATHEMATICAL MODEL OF THE THERMAL PERFORMANCE OF THE LINING OF THE CONVERTER’S WALLS, WHICH IS COVERED WITH SLAG, AT ITS INTERACTION WITH THE MELT DURING OXYGEN BLOWING Panteikov S.P., Panteikova E.S. The oxygen conve...
Моделювання в процесі прийняття рішення про придбання обладнання з врахуванням обмеження площі розміщення та розміру інвестицій
SIMULATION IN THE PROCESS OF MAKING DECISIONS ABOUT PURCHASING EQUIPMENT TAKING INTO ACCOUNT THE LIMITATIONS OF THE LOCATION AND OF THE SIZE THE INVESTMENT Karimov I.K., Karimov G.I. Abstract In the process of making...
МЕТОДИКА ФІЗИЧНОГО МОДЕЛЮВАННЯ СПЛИВАННЯ АНСАМБЛЮ НЕМЕТАЛЕВИХ ВКЛЮЧЕНЬ У СТАЛЕРОЗЛИВНОМУ КОВШІ
METHOD OF PHYSICAL MODELING OF FLOATING OF THE NONMETALLIC INCLUSIONS GROUP IN A TEEMING LADLE Lantukh O.S., Molchanov L.S., Synehin E.V. Abstract The authors considered an important problem for steel quality, consisti...
Математическое описание процесса преобразования измерительного сигнала в электронной системе
MATHEMATICAL DESCRIPTION OF PROCESS OF THE MEASURING SIGNAL SHAPING IS IN THE ELECTRONIC SYSTEM Meschaninov S.K., Bagriy V.V., Ustimenko V.O., Tonkonog E.A., Bogdanova E.N. Abstract Actuality. The reliability and accurac...
Комп’ютерна реалізація математичної моделі оптимального розміщення сучасних відділень зв'язку, що надають різні види послуг
The results of application of algorithms of decision of continuous nonlinear multigrocery problems of optimal set partition to the row of the applied tasks of the optimum placing of the modern communication offices that...