Solving the Ising Spin Glass Problem using a Bivariate EDA based on Markov Random Fields

Shakya, Siddhartha; McCall, John; Brown, Deryck F.

doi:10.1109/cec.2006.1688408

Cited by 20 publications

(30 citation statements)

References 19 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Though these experiments are constructed differently, the performance results are consistent with the experiments of [64]; exploiting the structure of spin glass systems is the key to solving them efficiently.…”

Section: Ising Spin Glassessupporting

confidence: 71%

MARLEDA: Effective distribution estimation through Markov random fields

Miikkulainen

Alden

2016

Theoretical Computer Science

View full text Add to dashboard Cite

Many problems within the biological sciences, such as DNA sequencing, protein structure prediction, and molecular docking, are being approached computationally. These problems require sophisticated solution methods that understand the complex natures of biological domains. Traditionally, such solution methods are problem specific, but recent advances in generic problem-solvers furnish hope for a new breed of computational tools. The challenge is to develop methods that can automatically learn or acquire an understanding of a complex problem domain.Estimation of Distribution Algorithms (EDAs) are generic search methods that use statistical models to learn the structure of a problem domain.EDAs have been successfully applied to many difficult search problems, such as circuit design, optimizing Ising spin glasses, and various scheduling tasks. Computational biology is a computational science that promises tangible, life-altering benefits. Research in fields such as genomics, protein structure prediction, and molecular docking, finds application in pharmacology and gene therapy. Consequently, day-to-day medicine and health have been improved, with the greatest breakthroughs still to come. These fields directly depend on advances in computer hardware and software, and there is great demand for effective computational problem-solving methods.Research in computational biology produces computational models of real-world objects or phenomena, such as molecular interactions or phylogenetic trees. The computational models can be studied, adjusted, and experimented with to generate predictions about their real-world counterparts. In 2 turn, real-world experiments help form the basis for new computational models. There is an interplay between real-world and computational research; each informs and refines the other. While a computational formulation of real-world problems is useful and convenient, there are drawbacks.The key difficulty of many computational problems, such as DNA sequence alignment, selecting key attributes for data mining, or optimizing antenna design, is that no analytic solution methods exist. There are also problems that do have analytic solutions, but those methods are too computationally expensive to be practical. Without the ability to construct an ideal solution efficiently, we are often reduced to a procedure of guess-and-check.Such a procedure, formally known as search, tries to identity the best solution(s) among a set of candidates by systematically evaluating alternatives.Search algorithms are, in many cases, a relatively easy and effective means of producing near-optimal solutions to difficult problems. However, in complex domains search is often inefficient, taking too long to be practical. This dissertation develops a search method that handles complexity better by statistically modeling the problem domain. ApproachFortunately, even when analytical methods are not available, for many computational problems it is relatively easy to test the "correctness" of a potential solution. For example, t...

show abstract

Section: Ising Spin Glassessupporting

confidence: 71%

MARLEDA: Effective distribution estimation through Markov random fields

Miikkulainen

Alden

2016

Theoretical Computer Science

View full text Add to dashboard Cite

show abstract

“…Previous publications on DEUM including [54,5] describe how a Markov network is used to model the distribution of energy across the set of variables in a bit-string encoded problem. In this section we summarise how the model is derived.…”

Section: Defining the Modelmentioning

confidence: 99%

“…Within DEUM, direct sampling is used to generate new solutions with a high probability of being high in fitness [54,53,5]. This direct sampling of the fitness model rather than the fitness function has the benefit that the model can make the problem easier for the search part of the algorithm -the smoothing effect described in [63].…”

Section: Applicationsmentioning

confidence: 99%

“…This direct sampling of the fitness model rather than the fitness function has the benefit that the model can make the problem easier for the search part of the algorithm -the smoothing effect described in [63]. For example, in [54], it was found that directly sampling solutions from the MFM rather than the true fitness function for 2D Ising increased the success rate of a bitwise Gibbs sampler from 87% to 99%. It was concluded that by using a real-valued search space within the MFM rather than the discrete values within the fitness function, the fitness landscape was altered to allow the algorithm to escape being trapped within plateaux, leading to more efficient searching of the landscape.…”

Section: Applicationsmentioning

confidence: 99%

See 1 more Smart Citation

The Markov Network Fitness Model

Brownlee

McCall

Shakya

2012

Adaptation, Learning, and Optimization

View full text Add to dashboard Cite

Fitness modelling is an area of research which has recently received much interest among the evolutionary computing community. Fitness models can improve the efficiency of optimisation through direct sampling to generate new solutions, guiding of traditional genetic operators or as surrogates for a noisy or long-running fitness functions. In this chapter we discuss the application of Markov networks to fitness modelling of black-box functions within evolutionary computation, accompanied by discussion on the relationship between Markov networks and Walsh analysis of fitness functions. We review alternative fitness modelling and approximation techniques and draw comparisons with the Markov network approach. We discuss the applicability of Markov networks as fitness surrogates which may be used for constructing guided operators or more general hybrid algorithms. We conclude with some observations and issues which arise from work conducted in this area so far.

show abstract

“…Many other examples exist in the literature including noisy objective functions [1], plateaus and multi-modality [22,18,19,26], deceptive gradient [3] and constraints [15]. Further examples of surrogate functions can be found in the extensive reviews by Jin [13,14].…”

Section: Examplesmentioning

confidence: 99%