The Experts below are selected from a list of 288 Experts worldwide ranked by ideXlab platform
Antonio Salmeron - One of the best experts on this subject based on the ideXlab platform.
-
Approximate Probability propagation with mixtures of truncated exponentials
European conference on symbolic and quantitative approaches to reasoning with uncertainty, 2007Co-Authors: Rafael Rumi, Antonio SalmeronAbstract:Mixtures of truncated exponentials (MTEs) are a powerful alternative to discretisation when working with hybrid Bayesian networks. One of the features of the MTE model is that standard propagation algorithms can be used. However, the complexity of the process is too high and therefore Approximate methods, which tradeoff complexity for accuracy, become necessary. In this paper we propose an Approximate propagation algorithm for MTE networks which is based on the Penniless propagation method already known for discrete variables. We also consider how to use Markov Chain Monte Carlo to carry out the Probability propagation. The performance of the proposed methods is analysed in a series of experiments with random networks.
-
algorithms for Approximate Probability propagation in bayesian networks
2004Co-Authors: Andres Cano, Serafin Moral, Antonio SalmeronAbstract:When a Bayesian network is defined over a very large or complicated domain, computing the posterior probabilities given some evidence may become unfeasible. In fact, Probability propagation is known to be an NP-hard problem. Since it is common to find huge domains in practical applications, Approximate algorithms have been developed. These algorithms compute estimations of the posterior probabilities with lower requirements, in terms of memory and computing time, than exact algorithms. In this paper we present some of the most recent developments in the area of Approximate propagation algorithms.
-
novel strategies to Approximate Probability trees in penniless propagation
International Journal of Intelligent Systems, 2003Co-Authors: Andres Cano, Serafin Moral, Antonio SalmeronAbstract:In this article we introduce some modifications over the Penniless propagation algorithm. When a message through the join tree is Approximated, the corresponding error is quantified in terms of an improved information measure, which leads to a new way of pruning several values in a Probability tree (representing a message) by a single one, computed from the value stored in the tree being pruned but taking into account the message stored in the opposite direction. Also, we have considered the possibility of replacing small Probability values by zero. Locally, this is not an optimal approximation strategy, but in Penniless propagation many different local approximations are carried out in order to estimate the posterior probabilities and, as we show in some experiments, replacing by zeros can improve the quality of the final approximations. © 2003 Wiley Periodicals, Inc.
J Trcebicki - One of the best experts on this subject based on the ideXlab platform.
-
Approximate Probability distributions for stochastic systems maximum entropy method
Computer Methods in Applied Mechanics and Engineering, 1999Co-Authors: K Sobczyk, J TrcebickiAbstract:Abstract The effective analytical methods for stochastic nonlinear dynamical systems are applicable only in some simple cases. If one deals with more complex systems and with the so-called real life applications the Approximate methods and numerical integration are necessary. In this paper we present the possible approaches to Approximate characterization of the Probability distributions of stochastic nonlinear systems. Starting from the description of the basic properties of such systems, the most notable recent efforts to evaluation of their Probability distributions are presented with emphasis on the maximum entropy method. This method, originated in its simple classical form in statistical physics, when suitably generalized, allows complicated stochastic systems to be treated successfully using information contained in the equations for statistical moments of the solution (or response). In this paper, the general scheme of the method is presented both for stationary and nonstationary distributions and then its numerical implementation is expounded. Nonlinear stochastic oscillatory systems are treated in detail and the obtained Probability distributions are shown graphically in comparison with the exact solutions and with the simulation results.
Andres Cano - One of the best experts on this subject based on the ideXlab platform.
-
algorithms for Approximate Probability propagation in bayesian networks
2004Co-Authors: Andres Cano, Serafin Moral, Antonio SalmeronAbstract:When a Bayesian network is defined over a very large or complicated domain, computing the posterior probabilities given some evidence may become unfeasible. In fact, Probability propagation is known to be an NP-hard problem. Since it is common to find huge domains in practical applications, Approximate algorithms have been developed. These algorithms compute estimations of the posterior probabilities with lower requirements, in terms of memory and computing time, than exact algorithms. In this paper we present some of the most recent developments in the area of Approximate propagation algorithms.
-
novel strategies to Approximate Probability trees in penniless propagation
International Journal of Intelligent Systems, 2003Co-Authors: Andres Cano, Serafin Moral, Antonio SalmeronAbstract:In this article we introduce some modifications over the Penniless propagation algorithm. When a message through the join tree is Approximated, the corresponding error is quantified in terms of an improved information measure, which leads to a new way of pruning several values in a Probability tree (representing a message) by a single one, computed from the value stored in the tree being pruned but taking into account the message stored in the opposite direction. Also, we have considered the possibility of replacing small Probability values by zero. Locally, this is not an optimal approximation strategy, but in Penniless propagation many different local approximations are carried out in order to estimate the posterior probabilities and, as we show in some experiments, replacing by zeros can improve the quality of the final approximations. © 2003 Wiley Periodicals, Inc.
Serafin Moral - One of the best experts on this subject based on the ideXlab platform.
-
ICEIS (2) - Partial Abductive Inference in Bayesian Networks by Using Probability Trees
Enterprise Information Systems V, 2020Co-Authors: Luis M. De Campos, José A. Gámez, Serafin MoralAbstract:The problem of partial abductive inference in Bayesian networks is, in general, more complex to solve than other inference problems as Probability/evidence propagation or total abduction. When join trees are used as the graphical structure over which propagation will be carried out, the problem can be decomposed into two stages: (1) to obtain a join tree containing only the variables included in the explanation set, and (2) to solve a total abduction problem over this new join tree. In De Campos et al. (2002a) different techniques are studied in order to approach this problem, obtaining as a result that not always the methods which obtain join trees with smaller size are also those requiring less CPU time during the propagation phase. In this work we propose to use (exact and Approximate) Probability trees as the basic data structure for the representation of the Probability distributions used during the propagation. ?From our experiments, we observe how the use of exact Probability trees improves the efficiency of the propagation. Besides, when using Approximate Probability trees the method obtains very good approximations and the required resources decrease considerably.
-
algorithms for Approximate Probability propagation in bayesian networks
2004Co-Authors: Andres Cano, Serafin Moral, Antonio SalmeronAbstract:When a Bayesian network is defined over a very large or complicated domain, computing the posterior probabilities given some evidence may become unfeasible. In fact, Probability propagation is known to be an NP-hard problem. Since it is common to find huge domains in practical applications, Approximate algorithms have been developed. These algorithms compute estimations of the posterior probabilities with lower requirements, in terms of memory and computing time, than exact algorithms. In this paper we present some of the most recent developments in the area of Approximate propagation algorithms.
-
novel strategies to Approximate Probability trees in penniless propagation
International Journal of Intelligent Systems, 2003Co-Authors: Andres Cano, Serafin Moral, Antonio SalmeronAbstract:In this article we introduce some modifications over the Penniless propagation algorithm. When a message through the join tree is Approximated, the corresponding error is quantified in terms of an improved information measure, which leads to a new way of pruning several values in a Probability tree (representing a message) by a single one, computed from the value stored in the tree being pruned but taking into account the message stored in the opposite direction. Also, we have considered the possibility of replacing small Probability values by zero. Locally, this is not an optimal approximation strategy, but in Penniless propagation many different local approximations are carried out in order to estimate the posterior probabilities and, as we show in some experiments, replacing by zeros can improve the quality of the final approximations. © 2003 Wiley Periodicals, Inc.
K Sobczyk - One of the best experts on this subject based on the ideXlab platform.
-
Approximate Probability distributions for stochastic systems maximum entropy method
Computer Methods in Applied Mechanics and Engineering, 1999Co-Authors: K Sobczyk, J TrcebickiAbstract:Abstract The effective analytical methods for stochastic nonlinear dynamical systems are applicable only in some simple cases. If one deals with more complex systems and with the so-called real life applications the Approximate methods and numerical integration are necessary. In this paper we present the possible approaches to Approximate characterization of the Probability distributions of stochastic nonlinear systems. Starting from the description of the basic properties of such systems, the most notable recent efforts to evaluation of their Probability distributions are presented with emphasis on the maximum entropy method. This method, originated in its simple classical form in statistical physics, when suitably generalized, allows complicated stochastic systems to be treated successfully using information contained in the equations for statistical moments of the solution (or response). In this paper, the general scheme of the method is presented both for stationary and nonstationary distributions and then its numerical implementation is expounded. Nonlinear stochastic oscillatory systems are treated in detail and the obtained Probability distributions are shown graphically in comparison with the exact solutions and with the simulation results.