Evolutionary Optimization With Markov Random Field Prior

Abstract

Abstract-This paper discusses an evolutionary algorithm in which the constituent variables of a solution are modeled by a Markov random field (MRF). We maintain a population of potential solutions at every generation and for each solution a fitness value is calculated. The evolution, however, is not achieved through genetic recombination. Instead, each variable in a solution will be updated by sampling from its probability distribution. According to the MRF prior, local exploitation is encoded in the conditional probabilities. For evolutionary exploration, we estimate the probabilities as fitness-weighted statistics. These two kinds of search are combined smoothly in our algorithm. We compare it with two representative algorithms [iterated conditional modes (ICM) and simulated annealing (SA)] on noisy and textured image segmentation. Remarkable performance is observed.Index Terms-Combinatorial optimization, estimation of distribution algorithms, evolutionary computation, genetic algorithms (GAs), hybrid search algorithm, image segmentation, Markov random fields (MRF).

On Cognitive Searching Optimization in Semi‐Markov Jump Decision Using Multistep Transition and Mental Rehearsal

“…e dimensional optimization method connects with the decision-making property such as the non-Markovian property [27], which is used to describe the cross-influence between different decision states. To reduce the dimension of decision searching, Engel et al [28] considered the stochastic jumpy interval in human cognitive decision behaviors and handled it with a linearity weighted logic according to monotonically increased time [29].…”

A review on probabilistic graphical models in evolutionary computation

“…Because of model learning complexity, Markov network-based EDAs (Santana 2003;Wang and Wang 2004;Shakya 2006;Alden 2007) are usually applied to applications where the structure of the optimization problem is known and can be easily represented using an undirected graphical model. However, an approximation of the probability distribution, like Kikuchi approximations (Santana 2005), can also be estimated to obtain the factorization of problem variables.…”

Research topics in discrete estimation of distribution algorithms based on factorizations

“…Among the alternatives proposed to deal with this question are: Partially evaluating one solution [67], reducing the number of individuals that are evaluated in each population [88,118] and estimating the fitness function using the models [76,83,103,105,106]. In other cases, factorizations constructed using message passing algorithms could be employed to estimate the function [31].…”

“…There is a class of hybrid EDAs [81,104,118] that uses the probabilistic models learned during the search to implement advanced local optimizers. Probabilistic modeling offer a variety of possibilities for the design of efficient local search strategies that use probabilistic models.…”