Publication detail

Modelling of Probability Distribution with Maximum Entropy

KARPÍŠEK, Z. JURÁK, P.

Czech title

Modelování rozdělení pravděpodobnosti s maximální entropií

English title

Modelling of Probability Distribution with Maximum Entropy

Type

conference paper

Language

en

Original abstract

The paper is concerned with the solution of a statistical problem of finding discrete probability distributions conforming to the requirement of maximum entropy under conditions given by estimates of their general moments from the observed relative frequencies. It is shown that the distributions derived are of an exponential type with maximum likelihood estimations of parameters that are also estimations by and modified chi square method. The new PC software Entropy for estimations is described and the results are illustrated by examples. Applications of the method are possible expect among others to control of the genetic algorithms and neural nets.

Czech abstract

Článek je zaměřen na řešení statistického problému nalezení diskrétního rozdělení pravděpodobnosti s maximální entropií za podmínek daných odhady jeho obecných momentů z pozorovaných relativních četností. Je ukázáno, že toto rozdělení je exponenciálního typu s maximálně věrohodnými odhady parametrů a také modifikovanou chí kvadrát metodou. Je také popsán nový PC software Entropy pro odhady a výsledky jsou ilustrovány na příkladech. Aplikace metody jsou možné mimo jiné při řízení genetických algoritmů a neuronových sítí.

English abstract

The paper is concerned with the solution of a statistical problem of finding discrete probability distributions conforming to the requirement of maximum entropy under conditions given by estimates of their general moments from the observed relative frequencies. It is shown that the distributions derived are of an exponential type with maximum likelihood estimations of parameters that are also estimations by and modified chi square method. The new PC software Entropy for estimations is described and the results are illustrated by examples. Applications of the method are possible expect among others to control of the genetic algorithms and neural nets.

Keywords in English

maximum entropy, distribution fitting, moment conditions, maximum likelihood estimate, modified minimum chi-squared method, PC software Entropy

RIV year

2001

Released

01.01.2001

Location

Brno University of Technology, Faculty of Mechanical Engineering, Brno

ISBN

80-214-1894-X

Book

MENDEL ´01. 7th International Conference on Soft Computing

Edition number

1

Pages count

8

BIBTEX


@inproceedings{BUT3838,
  author="Zdeněk {Karpíšek} and Petr {Jurák},
  title="Modelling of Probability Distribution with Maximum Entropy",
  booktitle="MENDEL ´01. 7th International Conference on Soft Computing",
  year="2001",
  month="January",
  address="Brno University of Technology, Faculty of Mechanical Engineering, Brno",
  isbn="80-214-1894-X"
}