2012/01/10 by Martin Pelikán, Pelikan, Martin, Mark W. Hauschild +1 · 1 citation
Computer Science · Engineering · #Artificial Intelligence (cs.AI) #Bayesian Modeling and Causal Inference #FOS: Computer and information sciences #Fault Detection and Control Systems #G.1.6 #I.2.6 #I.2.8 #Machine Learning and Algorithms #Machine Learning and Data Classification #Neural and Evolutionary Computing (cs.NE)
paper · pdf · doi:10.48550/arxiv.1201.2241
openalex publication_date 2012/01/10 · openalex created_date 2022/10/04 · openalex updated_date 2026/07/28
For many optimization problems it is possible to define a distance metric\nbetween problem variables that correlates with the likelihood and strength of\ninteractions between the variables. For example, one may define a metric so\nthat the dependencies between variables that are closer to each other with\nrespect to the metric are expected to be stronger than the dependencies between\nvariables that are further apart. The purpose of this paper is to describe a\nmethod that combines such a problem-specific distance metric with information\nmined from probabilistic models obtained in previous runs of estimation of\ndistribution algorithms with the goal of solving future problem instances of\nsimilar type with increased speed, accuracy and reliability. While the focus of\nthe paper is on additively decomposable problems and the hierarchical Bayesian\noptimization algorithm, it should be straightforward to generalize the approach\nto other model-directed optimization techniques and other problem classes.\nCompared to other techniques for learning from experience put forward in the\npast, the proposed technique is both more practical and more broadly\napplicable.\n