This chapter presents an entropy-based convergence measurement applicable to Estimation of Distri... more This chapter presents an entropy-based convergence measurement applicable to Estimation of Distribution Algorithms. Based on the measured entropy, the time point when the generation of new solutions becomes ineffective, can be detected. The proposed termination criterion is inherent to the complexity of used probabilistic models and automatically postpones the termination if inappropriate models are used.
The Bayesian Optimization Algorithm (BOA) belongs to the probabilistic model building genetic alg... more The Bayesian Optimization Algorithm (BOA) belongs to the probabilistic model building genetic algorithms where crossover and mutation operators are replaced by probability estimation and sampling techniques. The learned Bayesian network BN as the most general probability model is used to encode the structure of solved combinatorial problems. In [1] we proposed and simulated the pipeline hardware architecture for BOA. The aim of this paper is to propose the distributed version of BOA algorithm with a coarse-grained parallelism. We focused primarily on the construction of Bayesian network in the distributed environment. In adittion, methods for overlapping the communication latency during generation, evaluation and broadcasting of new population among the processes are described. Much attention was devoted to the implementation of proposed approaches using a cluster of workstations as a computational platform.
This paper deals with the k-way ratio cut hypergraph partitioning utilizing the Mixed discrete co... more This paper deals with the k-way ratio cut hypergraph partitioning utilizing the Mixed discrete continuous variant of the Bayesian Optimization Algorithm (mBOA). We have tested our algorithm on three partitioning taxonomies: recursive minimum ratio cut, multi-way minimum ratio cut and recursive minimum cut bisection. We have also derived a new approach for modeling of Boolean functions using binary decision diagrams (BDDs) which are primarily used as a probabilistic model of the mBOA algorithm.
... n denotes the number of binary decision variables and m the number of objectives and constrai... more ... n denotes the number of binary decision variables and m the number of objectives and constraints. ... 4.2 Comparison with Other MOEAs In order to evaluate BMOA with respect to other MOEAs we use the results of the comparative case study from [10] and focus on the large ...
The paper summarizes our recent work on the design, analysis and applications of the Bayesian opt... more The paper summarizes our recent work on the design, analysis and applications of the Bayesian optimization algorithm (BOA) and its advanced accelerated variants for solving complex – sometimes NP-complete – combinatorial optimization problems from circuit design. We review the methods for accelerating BOA for hypergraph-partitioning problem. The first method accelerates the convergence of sequential BOA by utilizing specific knowledge about the optimized problem and the second method is based on the parallel construction of a probabilistic model. In the experimental part we analyze the advantages of acceleration techniques and prove that BOA is able to solve hypergraph partitioning problems reliably, effectively, and without the need for specifying control parameters and encoding schemes as in recombination-based genetic algorithms.
This paper presents a hybrid evolutionary optimization strategy combining the Mixed Bayesian Opti... more This paper presents a hybrid evolutionary optimization strategy combining the Mixed Bayesian Optimization Algorithm (MBOA) with variance adaptation as implemented in Evolution Strategies. This new approach is intended to circumvent some of the deficiences of MBOA with unimodal functions and to enhance its adaptivity. The Adaptive MBOA algorithm–AMBOA–is compared with the Covariance Matrix Adaptation Evolution Strategy (CMA-ES). The comparison shows that, in continuous domains, AMBOA is more efficient ...
One of the most important challenges in computational optimization is the design of advanced blac... more One of the most important challenges in computational optimization is the design of advanced black-box optimization techniques that would enable automated, robust, and scalable solution to challenging optimization problems. This paper describes an advanced black-box optimizer—the hierarchical Bayesian optimization algorithm (hBOA)—that combines techniques of genetic and evolutionary computation, machine learning, and statistics to create a widely applicable tool for solving
ABSTRACT This paper is an experimental study on an utilization of additional knowledge about the ... more ABSTRACT This paper is an experimental study on an utilization of additional knowledge about the decomposition problem to be solved. We have demonstrated this approach on the hypergraph bisectioning that can serve as a model of system decomposition in common, data base decomposition etc. We have focused on the extension of the Bayesian Optimization Algorithm BOA. The extension of the original BOA algorithm is based on the usage of a prior information about the hypergraph structure. This knowledge is used for both setting initial Bayesian network and the initial population using injection of clusters to improve the convergence of the decomposition process. The behaviour of our version KBOA is tested on the set of benchmarks, such as grid and random geometric graphs as well as real hypergraphs.
This paper is an experimental study investigating the capability of Bayesian optimization algorit... more This paper is an experimental study investigating the capability of Bayesian optimization algorithms to solve dynamic problems. We tested the performance of two variants of Bayesian optimization algorithms – Mixed continuous-discrete Bayesian Optimization Algorithm (MBOA), Adaptive Mixed Bayesian Optimization Algorithm (AMBOA) – and new proposed modifications with embedded Sentinels concept and Hypervariance. We have compared the performance of these variants
ABSTRACT In the last few years there has been a growing interest in the field of Estimation of Di... more ABSTRACT In the last few years there has been a growing interest in the field of Estimation of Distribution Algorithms (EDAs), where crossover and mutation genetic operators are replaced by probability estimation and sampling techniques. The Bayesian Optimization Algorithm incorporates methods for learning Bayesian networks and uses these to model the promising solutions and generate new ones. The aim of this paper is to propose the parallel version of this algorithm, where the optimization time decreases linearly with the number of processors. During the parallel construction of network, the explicit topological ordering of variables is used to keep the model acyclic. The performance of the optimization process seems to be not affected by this constraint and our version of algorithm was successfully tested for the discrete combinatorial problem represented by graph partitioning as well as for deceptive functions.
Abstract: This paper is an experimental ,study on hypegraph ,partitioning using ,the simple genet... more Abstract: This paper is an experimental ,study on hypegraph ,partitioning using ,the simple genetic algorithm (GA) based on the ,schema ,theorem,and ,the advanced ,algorithms based ,on the ,estimation of distribution ,of promising,solution. Primarily we have implemented,a simple,GA based on the,GaLib library[Gal94] and some hybrid variant included a fast heuristics to speed up the convergence,of the optimization process. Secondly we have,implemented
... Binary decision trees have been successfully used in EDA [2] for discrete optimization proble... more ... Binary decision trees have been successfully used in EDA [2] for discrete optimization problems-with ... define the elementary models for obtaining the" target" variable X,. For continuous variables one ... In the present version we use more sophisticated metrics to find the optimal split ...
Mass spectrometry data generated in differential profiling of complex protein samples are classic... more Mass spectrometry data generated in differential profiling of complex protein samples are classically exploited using database searches. In addition, quantitative profiling is performed by various methods, one of them using isotopically coded affinity tags, where one typically uses a light and a heavy tag. Here, we present a new algorithm, ICATcher, which detects pairs of light/heavy peptide MS/MS spectra independent of sequence databases. The method can be used for de novo sequencing and detection of posttranslational modifications. ICATcher is distributed as open source software.
This chapter presents an entropy-based convergence measurement applicable to Estimation of Distri... more This chapter presents an entropy-based convergence measurement applicable to Estimation of Distribution Algorithms. Based on the measured entropy, the time point when the generation of new solutions becomes ineffective, can be detected. The proposed termination criterion is inherent to the complexity of used probabilistic models and automatically postpones the termination if inappropriate models are used.
The Bayesian Optimization Algorithm (BOA) belongs to the probabilistic model building genetic alg... more The Bayesian Optimization Algorithm (BOA) belongs to the probabilistic model building genetic algorithms where crossover and mutation operators are replaced by probability estimation and sampling techniques. The learned Bayesian network BN as the most general probability model is used to encode the structure of solved combinatorial problems. In [1] we proposed and simulated the pipeline hardware architecture for BOA. The aim of this paper is to propose the distributed version of BOA algorithm with a coarse-grained parallelism. We focused primarily on the construction of Bayesian network in the distributed environment. In adittion, methods for overlapping the communication latency during generation, evaluation and broadcasting of new population among the processes are described. Much attention was devoted to the implementation of proposed approaches using a cluster of workstations as a computational platform.
This paper deals with the k-way ratio cut hypergraph partitioning utilizing the Mixed discrete co... more This paper deals with the k-way ratio cut hypergraph partitioning utilizing the Mixed discrete continuous variant of the Bayesian Optimization Algorithm (mBOA). We have tested our algorithm on three partitioning taxonomies: recursive minimum ratio cut, multi-way minimum ratio cut and recursive minimum cut bisection. We have also derived a new approach for modeling of Boolean functions using binary decision diagrams (BDDs) which are primarily used as a probabilistic model of the mBOA algorithm.
... n denotes the number of binary decision variables and m the number of objectives and constrai... more ... n denotes the number of binary decision variables and m the number of objectives and constraints. ... 4.2 Comparison with Other MOEAs In order to evaluate BMOA with respect to other MOEAs we use the results of the comparative case study from [10] and focus on the large ...
The paper summarizes our recent work on the design, analysis and applications of the Bayesian opt... more The paper summarizes our recent work on the design, analysis and applications of the Bayesian optimization algorithm (BOA) and its advanced accelerated variants for solving complex – sometimes NP-complete – combinatorial optimization problems from circuit design. We review the methods for accelerating BOA for hypergraph-partitioning problem. The first method accelerates the convergence of sequential BOA by utilizing specific knowledge about the optimized problem and the second method is based on the parallel construction of a probabilistic model. In the experimental part we analyze the advantages of acceleration techniques and prove that BOA is able to solve hypergraph partitioning problems reliably, effectively, and without the need for specifying control parameters and encoding schemes as in recombination-based genetic algorithms.
This paper presents a hybrid evolutionary optimization strategy combining the Mixed Bayesian Opti... more This paper presents a hybrid evolutionary optimization strategy combining the Mixed Bayesian Optimization Algorithm (MBOA) with variance adaptation as implemented in Evolution Strategies. This new approach is intended to circumvent some of the deficiences of MBOA with unimodal functions and to enhance its adaptivity. The Adaptive MBOA algorithm–AMBOA–is compared with the Covariance Matrix Adaptation Evolution Strategy (CMA-ES). The comparison shows that, in continuous domains, AMBOA is more efficient ...
One of the most important challenges in computational optimization is the design of advanced blac... more One of the most important challenges in computational optimization is the design of advanced black-box optimization techniques that would enable automated, robust, and scalable solution to challenging optimization problems. This paper describes an advanced black-box optimizer—the hierarchical Bayesian optimization algorithm (hBOA)—that combines techniques of genetic and evolutionary computation, machine learning, and statistics to create a widely applicable tool for solving
ABSTRACT This paper is an experimental study on an utilization of additional knowledge about the ... more ABSTRACT This paper is an experimental study on an utilization of additional knowledge about the decomposition problem to be solved. We have demonstrated this approach on the hypergraph bisectioning that can serve as a model of system decomposition in common, data base decomposition etc. We have focused on the extension of the Bayesian Optimization Algorithm BOA. The extension of the original BOA algorithm is based on the usage of a prior information about the hypergraph structure. This knowledge is used for both setting initial Bayesian network and the initial population using injection of clusters to improve the convergence of the decomposition process. The behaviour of our version KBOA is tested on the set of benchmarks, such as grid and random geometric graphs as well as real hypergraphs.
This paper is an experimental study investigating the capability of Bayesian optimization algorit... more This paper is an experimental study investigating the capability of Bayesian optimization algorithms to solve dynamic problems. We tested the performance of two variants of Bayesian optimization algorithms – Mixed continuous-discrete Bayesian Optimization Algorithm (MBOA), Adaptive Mixed Bayesian Optimization Algorithm (AMBOA) – and new proposed modifications with embedded Sentinels concept and Hypervariance. We have compared the performance of these variants
ABSTRACT In the last few years there has been a growing interest in the field of Estimation of Di... more ABSTRACT In the last few years there has been a growing interest in the field of Estimation of Distribution Algorithms (EDAs), where crossover and mutation genetic operators are replaced by probability estimation and sampling techniques. The Bayesian Optimization Algorithm incorporates methods for learning Bayesian networks and uses these to model the promising solutions and generate new ones. The aim of this paper is to propose the parallel version of this algorithm, where the optimization time decreases linearly with the number of processors. During the parallel construction of network, the explicit topological ordering of variables is used to keep the model acyclic. The performance of the optimization process seems to be not affected by this constraint and our version of algorithm was successfully tested for the discrete combinatorial problem represented by graph partitioning as well as for deceptive functions.
Abstract: This paper is an experimental ,study on hypegraph ,partitioning using ,the simple genet... more Abstract: This paper is an experimental ,study on hypegraph ,partitioning using ,the simple genetic algorithm (GA) based on the ,schema ,theorem,and ,the advanced ,algorithms based ,on the ,estimation of distribution ,of promising,solution. Primarily we have implemented,a simple,GA based on the,GaLib library[Gal94] and some hybrid variant included a fast heuristics to speed up the convergence,of the optimization process. Secondly we have,implemented
... Binary decision trees have been successfully used in EDA [2] for discrete optimization proble... more ... Binary decision trees have been successfully used in EDA [2] for discrete optimization problems-with ... define the elementary models for obtaining the" target" variable X,. For continuous variables one ... In the present version we use more sophisticated metrics to find the optimal split ...
Mass spectrometry data generated in differential profiling of complex protein samples are classic... more Mass spectrometry data generated in differential profiling of complex protein samples are classically exploited using database searches. In addition, quantitative profiling is performed by various methods, one of them using isotopically coded affinity tags, where one typically uses a light and a heavy tag. Here, we present a new algorithm, ICATcher, which detects pairs of light/heavy peptide MS/MS spectra independent of sequence databases. The method can be used for de novo sequencing and detection of posttranslational modifications. ICATcher is distributed as open source software.
Uploads
Papers by Jiri Ocenasek