Conditional problem for objective probability

Otakar Kříž

Kybernetika (1998)

  • Volume: 34, Issue: 1, page [27]-40
  • ISSN: 0023-5954

Abstract

top
Marginal problem (see [Kel]) consists in finding a joint distribution whose marginals are equal to the given less-dimensional distributions. Let’s generalize the problem so that there are given not only less-dimensional distributions but also conditional probabilities. It is necessary to distinguish between objective (Kolmogorov) probability and subjective (de Finetti) approach ([Col,Sco]). In the latter, the coherence problem incorporates both probabilities and conditional probabilities in a unified framework. Different algorithms available for its solution are described e. g. in ([Gil,Col,Vic]). In the context of the former approach, it will be shown that it is possible to split the task into solving the marginal problem independently and to subsequent solving pure “conditional" problem as certain type of optimization. First, an algorithm (Conditional problem) that generates a distribution whose conditional probabilities are equal to the given ones is presented. Due to the multimodality of the criterion function, the algorithm is only heuristical. Due to the computational complexity, it is efficient for small size problems e. g. 5 dichotomical variables. Second, a method is mentioned how to unite marginal and conditional problem to a more general consistency problem for objective probability. Due to computational complexity, both algorithms are effective only for limited number of variables and conditionals. The described approach makes possible to integrate in the solution of the consistency problem additional knowledge contained e. g. in an empirical distribution.

How to cite

top

Kříž, Otakar. "Conditional problem for objective probability." Kybernetika 34.1 (1998): [27]-40. <http://eudml.org/doc/33332>.

@article{Kříž1998,
abstract = {Marginal problem (see [Kel]) consists in finding a joint distribution whose marginals are equal to the given less-dimensional distributions. Let’s generalize the problem so that there are given not only less-dimensional distributions but also conditional probabilities. It is necessary to distinguish between objective (Kolmogorov) probability and subjective (de Finetti) approach ([Col,Sco]). In the latter, the coherence problem incorporates both probabilities and conditional probabilities in a unified framework. Different algorithms available for its solution are described e. g. in ([Gil,Col,Vic]). In the context of the former approach, it will be shown that it is possible to split the task into solving the marginal problem independently and to subsequent solving pure “conditional" problem as certain type of optimization. First, an algorithm (Conditional problem) that generates a distribution whose conditional probabilities are equal to the given ones is presented. Due to the multimodality of the criterion function, the algorithm is only heuristical. Due to the computational complexity, it is efficient for small size problems e. g. 5 dichotomical variables. Second, a method is mentioned how to unite marginal and conditional problem to a more general consistency problem for objective probability. Due to computational complexity, both algorithms are effective only for limited number of variables and conditionals. The described approach makes possible to integrate in the solution of the consistency problem additional knowledge contained e. g. in an empirical distribution.},
author = {Kříž, Otakar},
journal = {Kybernetika},
keywords = {marginal problem; algorithm; marginal problem; algorithm},
language = {eng},
number = {1},
pages = {[27]-40},
publisher = {Institute of Information Theory and Automation AS CR},
title = {Conditional problem for objective probability},
url = {http://eudml.org/doc/33332},
volume = {34},
year = {1998},
}

TY - JOUR
AU - Kříž, Otakar
TI - Conditional problem for objective probability
JO - Kybernetika
PY - 1998
PB - Institute of Information Theory and Automation AS CR
VL - 34
IS - 1
SP - [27]
EP - 40
AB - Marginal problem (see [Kel]) consists in finding a joint distribution whose marginals are equal to the given less-dimensional distributions. Let’s generalize the problem so that there are given not only less-dimensional distributions but also conditional probabilities. It is necessary to distinguish between objective (Kolmogorov) probability and subjective (de Finetti) approach ([Col,Sco]). In the latter, the coherence problem incorporates both probabilities and conditional probabilities in a unified framework. Different algorithms available for its solution are described e. g. in ([Gil,Col,Vic]). In the context of the former approach, it will be shown that it is possible to split the task into solving the marginal problem independently and to subsequent solving pure “conditional" problem as certain type of optimization. First, an algorithm (Conditional problem) that generates a distribution whose conditional probabilities are equal to the given ones is presented. Due to the multimodality of the criterion function, the algorithm is only heuristical. Due to the computational complexity, it is efficient for small size problems e. g. 5 dichotomical variables. Second, a method is mentioned how to unite marginal and conditional problem to a more general consistency problem for objective probability. Due to computational complexity, both algorithms are effective only for limited number of variables and conditionals. The described approach makes possible to integrate in the solution of the consistency problem additional knowledge contained e. g. in an empirical distribution.
LA - eng
KW - marginal problem; algorithm; marginal problem; algorithm
UR - http://eudml.org/doc/33332
ER -

References

top
  1. Cheeseman P., A method of computing generalized Bayesian probability values of expert systems, In: Proceedings of the 6-th Joint Conference on Artificial Intelligence (IJCAI-83), Karlsruhe, pp. 198–202 
  2. Deming W. E., Stephan F. F., 10.1214/aoms/1177731829, Ann. Math. Statist. 11 (1940), 427–444 (1940) MR0003527DOI10.1214/aoms/1177731829
  3. Gilio A., Ingrassia S., Geometrical aspects in checking coherence of probability assessments, In: IPMU’96: Proceedings of the 6th International IPMU Conference (B. Bouchon–Meunier, M. Delgado, J. L. Verdegay, M. A. Vila, R. Yager, eds.), Granada 1996, pp. 55–59 (1996) 
  4. Coletti G., Scozzafava R., 10.1142/S021848859600007X, Internat. J. Uncertainty, Fuzziness and Knowledge–Based Systems, 4 (1996), 2, 103–127 (1996) Zbl1232.03010MR1390898DOI10.1142/S021848859600007X
  5. Kellerer H. G., 10.1007/BF00534912, Z. Wahrsch. verw. Gebiete 3 (1964), 247–270 (1964) Zbl0126.34003MR0175158DOI10.1007/BF00534912
  6. Kříž O., Invariant moves for constructing extensions of marginals, In: IPMU’94: Proceedings of the 5th International IPMU Conference (B. Bouchon–Meunier, R. Yager, eds.), Paris 1994, pp. 984–989 (1994) 
  7. Kříž O., Optimizations on finite–dimensional distributions with fixed marginals, In: WUPES 94: Proceedings of the 3-rd Workshop on Uncertainty Processing (R. Jiroušek, ed.), Třešť 1994, pp. 143–156 (1994) 
  8. Kříž O., Marginal problem on finite sets, In: IPMU’96: Proceedings of the 6-th International IPMU Conference (B. Bouchon–Meunier, M. Delgado, J. L. Verdegay, M. A. Vila, R. Yager, eds.), Granada 1996, Vol. II, pp. 763–768 (1996) 
  9. Kříž O., Inconsistent marginal problem on finite sets, In: Distributions with Given Marginals and Moment Problems (J. Štěpán, V. Beneš, eds.), Kluwer Academic Publishers, Dordrecht – Boston – London 1997, pp. 235–242 (1997) Zbl0907.60003
  10. Scozzafava R., 10.1080/03043799508923366, European J. Engineering Education 20 (1995), 3, 353–363 (1995) DOI10.1080/03043799508923366
  11. Vicig P., An algorithm for imprecise conditional probability assesment in expert systems, In: IPMU’96: Proceedings of the 6-th International IPMU Conference (B. Bouchon–Meunier, M. Delgado, J. L. Verdegay, M. A. Vila, R. Yager, eds.), Granada, 1996, Vol. I, pp. 61–66 (1996) 

NotesEmbed ?

top

You must be logged in to post comments.

To embed these notes on your page include the following JavaScript code on your page where you want the notes to appear.

Only the controls for the widget will be shown in your chosen language. Notes will be shown in their authored language.

Tells the widget how many notes to show per page. You can cycle through additional notes using the next and previous controls.

    
                

Note: Best practice suggests putting the JavaScript code just before the closing </body> tag.