Model accuracy in the Bayesian optimization algorithm

Lima, Cláudio F.; Lobo, Fernando G.; Pelikán, Martin; Goldberg, David E.

doi:10.1007/s00500-010-0675-y

Cited by 25 publications

(10 citation statements)

References 48 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Indeed, including such inessential dependencies in the problem model used by the algorithm can hamper performance [2,19,23]. This concurs with our findings for the linkage-aware ILS when applied to the feature selection problem studied in our experiments.…”

Section: Linkage In Eassupporting

confidence: 82%

Evolutionary Algorithms with Linkage Information for Feature Selection in Brain Computer Interfaces

Adair

Brownlee

Ochoa

2016

Advances in Intelligent Systems and Computing

View full text Add to dashboard Cite

Brain Computer Interfaces are an essential technology for the advancement of prosthetic limbs, but current signal acquisition methods are hindered by a number of factors, not least, noise. In this context, Feature Selection is required to choose the important signal features and improve classifier accuracy. Evolutionary algorithms have proven to outperform filtering methods (in terms of accuracy) for Feature Selection. This paper applies a single-point heuristic search method, Iterated Local Search (ILS), and compares it to a genetic algorithm (GA) and a memetic algorithm (MA). It then further attempts to utilise Linkage between features to guide search operators in the algorithms stated. The GA was found to outperform ILS. Counter-intuitively, linkage-guided algorithms resulted in higher classification error rates than their unguided alternatives. Explanations for this are explored.

show abstract

Section: Linkage In Eassupporting

confidence: 82%

Evolutionary Algorithms with Linkage Information for Feature Selection in Brain Computer Interfaces

Adair

Brownlee

Ochoa

2016

Advances in Intelligent Systems and Computing

View full text Add to dashboard Cite

show abstract

“…We argue that minimal problem structure must be detected for efficient solution of the problem, and that detection of only the minimal structure further increases efficiency. It has already been established that spurious dependencies can decrease efficiency [11], [12] and that omission of parts of problem structure from the algorithm structure leads to reduced ability to rank solutions [5]. In EDAs, reduced structure means fewer model parameters to estimate, leading to reduced modelbuilding effort.…”

Section: Walsh Structuresmentioning

confidence: 99%

“…Common selection operators for this purpose are tournament selection and truncation selection, detailed further below. Poor configuration of the selection operator can lead to missing key interactions, or falsely detecting unnecessary ones, which both impact on performance [11], [12], [14].…”

Section: Structure Learning Approachesmentioning

confidence: 99%

“…Also, effort spent in learning non-spurious but inessential structure is wasted and may be quite costly. Recent works including [11], [12] have explored how such errors influence the scalability and efficiency of EDAs, the latter showing that spurious dependencies decrease performance. A much earlier work with EDAs based on directed models [13] showed that dependencies generated by selection may genuinely reflect such a distribution or may simply be due to a sampling bias imposed by the selection.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Structural coherence of problem and algorithm: An analysis for EDAs on all 2-bit and 3-bit problems

Brownlee

McCall

Christie

2015

2015 IEEE Congress on Evolutionary Computation (CEC)

View full text Add to dashboard Cite

This is an author produced version of a paper published in IEEE Congress on Evolutionary Computation (CEC): Proceedings (ISBN 9781479974924)This version may not include final proof corrections and does not include published layout or pagination. Citation DetailsCitation for the version of the work held in 'OpenAIR@RGU': Copyright Items in 'OpenAIR@RGU', Robert Gordon University Open Access Institutional Repository, are protected by copyright and intellectual property law. If you believe that any material held in 'OpenAIR@RGU' infringes copyright, please contact openair-help@rgu.ac.uk with details. The item will be removed from the repository while the claim is investigated. Abstract-Metaheuristics assume some kind of coherence between decision and objective spaces. Estimation of Distribution algorithms approach this by constructing an explicit probabilistic model of high fitness solutions, the structure of which is intended to reflect the structure of the problem. In this context, "structure" means the dependencies or interactions between problem variables in a probabilistic graphical model. There are many approaches to discovering these dependencies, and existing work has already shown that often these approaches discover "unnecessary" elements of structure -that is, elements which are not needed to correctly rank solutions. This work performs an exhaustive analysis of all 2 and 3 bit problems, grouped into classes based on mononotic invariance. It is shown in [1] that each class has a minimal Walsh structure that can be used to solve the problem. We compare the structure discovered by different structure learning approaches to the minimal Walsh structure for each class, with summaries of which interactions are (in)correctly identified. Our analysis reveals a large number of symmetries that may be used to simplify problem solving. We show that negative selection can result in improved coherence between discovered and necessary structure, and conclude with some directions for a general programme of study building on this work.for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works.

show abstract

“…In hBOA and other EDAs based on complex probabilistic models, building an accurate probabilistic model is crucial to the success [2,3,11,19]. However, building complex probabilistic models can be time consuming and it may require rather large populations of solutions [2,3].…”

Section: Learning From Experience Using Distance-based Biasmentioning

confidence: 99%

Transfer Learning, Soft Distance-Based Bias, and the Hierarchical BOA

Pelikán

Hauschild

Lanzi

2012

Lecture Notes in Computer Science

Self Cite

View full text Add to dashboard Cite

An automated technique has recently been proposed to transfer learning in the hierarchical Bayesian optimization algorithm (hBOA) based on distance-based statistics. The technique enables practitioners to improve hBOA efficiency by collecting statistics from probabilistic models obtained in previous hBOA runs and using the obtained statistics to bias future hBOA runs on similar problems. The purpose of this paper is threefold: (1) test the technique on several classes of NP-complete problems, including MAXSAT, spin glasses and minimum vertex cover; (2) demonstrate that the technique is effective even when previous runs were done on problems of different size; (3) provide empirical evidence that combining transfer learning with other efficiency enhancement techniques can often yield nearly multiplicative speedups. KeywordsTransfer learning, inductive transfer, learning from experience, estimation of distribution algorithms, hierarchical Bayesian optimization algorithm, decomposable problems, efficiency enhancement. Abstract An automated technique has recently been proposed to transfer learning in the hierarchical Bayesian optimization algorithm (hBOA) based on distance-based statistics. The technique enables practitioners to improve hBOA efficiency by collecting statistics from probabilistic models obtained in previous hBOA runs and using the obtained statistics to bias future hBOA runs on similar problems. The purpose of this paper is threefold: (1) test the technique on several classes of NP-complete problems, including MAXSAT, spin glasses and minimum vertex cover; (2) demonstrate that the technique is effective even when previous runs were done on problems of different size; (3) provide empirical evidence that combining transfer learning with other efficiency enhancement techniques can often yield nearly multiplicative speedups.

show abstract

Model accuracy in the Bayesian optimization algorithm

Cited by 25 publications

References 48 publications

Evolutionary Algorithms with Linkage Information for Feature Selection in Brain Computer Interfaces

Evolutionary Algorithms with Linkage Information for Feature Selection in Brain Computer Interfaces

Structural coherence of problem and algorithm: An analysis for EDAs on all 2-bit and 3-bit problems

Transfer Learning, Soft Distance-Based Bias, and the Hierarchical BOA

Contact Info

Product

Resources

About