Search

The paper provides an axiomatic setup for an entropy function as a measure of diversity. A general definition of cross entropy is given and its use in solving a variety of stochastic and nonstochastic optimization problems is mentioned. A method of deriving a cross entropy function associated with a given entropy function is given. [ABSTRACT FROM AUTHOR]

Image entropy thresholding approach has drawn the attentions in image segmentation. The endeavor of this paper is focused on multilevel thresholding using the minimum cross entropy criterion. In the literature, the particle swarm optimization (PSO) had been applied to conducting the threshold selection. The adopted algorithm used in this paper is the honey bee mating optimization (HBMO). In experiments, the two different methods are implemented for comparison with the results of segmentation. Compared to the results of PSO, the threshold selection of HBMO is more close to the optimal ones examined by the exhaustive search method. Furthermore, the segmentation result of HBMO is superior to PSO method, but it still slower than ones of PSO. [ABSTRACT FROM AUTHOR]

In this paper, cross entropy method is used for solving dual Lagrange support vector machine (SVM). Cross entropy (CE) method is a new practical approach which is widely used in some applications such as combinatorial optimization, learning algorithm and simulation. Our approach refers to Kernel Adatron which is solving dual Lagrange SVM using gradient ascent method. Hereby, the cross entropy method is applied to solve dual Lagrange SVM optimization problem to find the optimal or at least near optimal Lagrange multipliers as a solution. As known, the standard SVM with quadratic programming solver suffers from high computational time. Some real world datasets are used to test the algorithms and compare to the existing approach in terms of computation time and accuracy. Our approach is fast and produce good results in terms of generalization error. [ABSTRACT FROM AUTHOR]

In this paper we analyze the performance of three algorithms for soft evidential update, in which a probability distribution represented by a Bayesian network is modified to a new distribution constrained by given marginals, and closest to the original distribution according to cross entropy. The first algorithm is a new and improved version of the big clique algorithm [1] that utilizes lazy propagation [2]. The second and third algorithm [3] are wrapper methods that convert soft evidence to virtual evidence, in which the evidence for a variable consists of a likelihood ratio. Virtual evidential update is supported in existing Bayesian inference engines, such as Hugin. To evaluate the three algorithms, we implemented BRUSE (Bayesian Reasoning Using Soft Evidence), a new Bayesian inference engine, and instrumented it. The resulting statistics are presented and discussed. [ABSTRACT FROM AUTHOR]