Pub Date : 2013-09-08DOI: 10.1109/BRICS-CCI-CBIC.2013.105
Alisson Marques da Silva, A. Faria, Thiago de Souza Rodrigues, Marcelo Azevedo Costa, A. de Pádua Braga
Acute leukemia classification into its Myeloid and Lymphoblastic subtypes is usually accomplished according to the morphological appearance of the tumor. Nevertheless, cells from the two subtypes can have similar histopathological appearance, which makes screening procedures very difficult. Correct classification of patients in the initial phases of the disease would allow doctors to properly prescribe cancer treatment. Therefore, the development of alternative methods, to the usual morphological classification, is needed in order to improve classification rates and treatment. This paper is based on the principle that DNA microarray data extracted from tumors contain sufficient information to differentiate leukemia subtypes. The classification task is described as a general pattern recognition problem, requiring initial representation by causal quantitative features, followed by the construction of a classifier. In order to show the validity of our methods, a publicly available dataset of acute leukemia comprising 72 samples with 7,129 features was used. The dataset was split into two subsets: the training dataset with 38 samples and the test dataset with 34 samples. Feature selection methods were applied to the training dataset. The 50 most predictive genes, according to each method, were selected. Artificial Neural Network (ANN) classifiers were developed to compare the feature selection methods. Among the 50 genes selected using the best classifier, 21 are consistent with previous work and 4 additional ones are clearly related to tumor molecular processes. The remaining 25 selected genes were able to classify the test dataset, correctly, using the ANN.
{"title":"Artificial Neural Networks and Ranking Approach for Probe Selection and Classification of Microarray Data","authors":"Alisson Marques da Silva, A. Faria, Thiago de Souza Rodrigues, Marcelo Azevedo Costa, A. de Pádua Braga","doi":"10.1109/BRICS-CCI-CBIC.2013.105","DOIUrl":"https://doi.org/10.1109/BRICS-CCI-CBIC.2013.105","url":null,"abstract":"Acute leukemia classification into its Myeloid and Lymphoblastic subtypes is usually accomplished according to the morphological appearance of the tumor. Nevertheless, cells from the two subtypes can have similar histopathological appearance, which makes screening procedures very difficult. Correct classification of patients in the initial phases of the disease would allow doctors to properly prescribe cancer treatment. Therefore, the development of alternative methods, to the usual morphological classification, is needed in order to improve classification rates and treatment. This paper is based on the principle that DNA microarray data extracted from tumors contain sufficient information to differentiate leukemia subtypes. The classification task is described as a general pattern recognition problem, requiring initial representation by causal quantitative features, followed by the construction of a classifier. In order to show the validity of our methods, a publicly available dataset of acute leukemia comprising 72 samples with 7,129 features was used. The dataset was split into two subsets: the training dataset with 38 samples and the test dataset with 34 samples. Feature selection methods were applied to the training dataset. The 50 most predictive genes, according to each method, were selected. Artificial Neural Network (ANN) classifiers were developed to compare the feature selection methods. Among the 50 genes selected using the best classifier, 21 are consistent with previous work and 4 additional ones are clearly related to tumor molecular processes. The remaining 25 selected genes were able to classify the test dataset, correctly, using the ANN.","PeriodicalId":306195,"journal":{"name":"2013 BRICS Congress on Computational Intelligence and 11th Brazilian Congress on Computational Intelligence","volume":"2010 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2013-09-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131594266","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2013-09-08DOI: 10.1109/BRICS-CCI-CBIC.2013.70
Jonas Krause, H. S. Lopes
This paper provides a brief description on how continuous algorithms can be applied to binary problems. Differential Evolution is the continuous algorithm studied and two versions of this algorithm are presented: the Binary Differential Evolution with a binary encoding and the Discretized Differential Evolution with a continuous encoding. Several discretization methods are presented and the most used method in literature is implemented for the solution discretization. Benchmarks with different complexity and search space sizes of the Multiple Knapsack Problem are used to compare the performance of each Differential Evolution algorithm presented and the Genetic Algorithm with binary encoding. Results suggest that continuous methods can be very efficient when discretized for binary spaces.
{"title":"A Comparison of Differential Evolution Algorithm with Binary and Continuous Encoding for the MKP","authors":"Jonas Krause, H. S. Lopes","doi":"10.1109/BRICS-CCI-CBIC.2013.70","DOIUrl":"https://doi.org/10.1109/BRICS-CCI-CBIC.2013.70","url":null,"abstract":"This paper provides a brief description on how continuous algorithms can be applied to binary problems. Differential Evolution is the continuous algorithm studied and two versions of this algorithm are presented: the Binary Differential Evolution with a binary encoding and the Discretized Differential Evolution with a continuous encoding. Several discretization methods are presented and the most used method in literature is implemented for the solution discretization. Benchmarks with different complexity and search space sizes of the Multiple Knapsack Problem are used to compare the performance of each Differential Evolution algorithm presented and the Genetic Algorithm with binary encoding. Results suggest that continuous methods can be very efficient when discretized for binary spaces.","PeriodicalId":306195,"journal":{"name":"2013 BRICS Congress on Computational Intelligence and 11th Brazilian Congress on Computational Intelligence","volume":"29 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2013-09-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130511979","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2013-09-08DOI: 10.1109/BRICS-CCI-CBIC.2013.104
C. Benítez, Rafael Stubs Parpinelli, H. S. Lopes
This paper applies a heterogeneous parallel ecology-inspired algorithm (pECO) to solve a complex problem from bioinformatics. The ecological-inspired algorithm represents a new perspective to develop cooperative evolutionary algorithms. Different algorithms are applied to compose the computational ecosystem in a heterogeneous model. The aim is to search low energy conformations for the Protein Structure Prediction problem, concerning the 3D-AB off-lattice model. Being a problem that demands a lot of computational effort, a parallel master-slave architecture is employed in order to allow the application of the computational ecosystem in a reasonable computing time. From the results, the pECO approach obtained the best conformation for the 13 amino-acid long sequence and competitive results for the other sequences.
{"title":"A Heterogeneous Parallel Ecologically-Inspired Approach Applied to the 3D-AB Off-Lattice Protein Structure Prediction Problem","authors":"C. Benítez, Rafael Stubs Parpinelli, H. S. Lopes","doi":"10.1109/BRICS-CCI-CBIC.2013.104","DOIUrl":"https://doi.org/10.1109/BRICS-CCI-CBIC.2013.104","url":null,"abstract":"This paper applies a heterogeneous parallel ecology-inspired algorithm (pECO) to solve a complex problem from bioinformatics. The ecological-inspired algorithm represents a new perspective to develop cooperative evolutionary algorithms. Different algorithms are applied to compose the computational ecosystem in a heterogeneous model. The aim is to search low energy conformations for the Protein Structure Prediction problem, concerning the 3D-AB off-lattice model. Being a problem that demands a lot of computational effort, a parallel master-slave architecture is employed in order to allow the application of the computational ecosystem in a reasonable computing time. From the results, the pECO approach obtained the best conformation for the 13 amino-acid long sequence and competitive results for the other sequences.","PeriodicalId":306195,"journal":{"name":"2013 BRICS Congress on Computational Intelligence and 11th Brazilian Congress on Computational Intelligence","volume":"114 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2013-09-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123160519","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2013-09-08DOI: 10.1109/BRICS-CCI-CBIC.2013.49
Mardé Helbig, A. Engelbrecht
Dynamic multi-objective optimisation (DMOO) entails solving optimisation problems with more than one objective, where at least one objective changes over time. Normally at least two of the objectives are in conflict with one another. Therefore, a single solution does not exist and the goal of an algorithm is to find for each environment a set of solutions that are both diverse and as close as possible to the optimal trade-off solution set. Solving dynamic multi-objective optimisation problems (DMOOPs) is not a trivial task, since the field of DMOO has many challenges. This paper highlights these challenges, namely the selection of benchmark functions and performance measures, the analyses of obtained results and selecting a preferred solution from the set of trade-off solutions. In addition, this paper discusses emerging research areas within computational intelligence (CI), such as hyper-heuristics, constrained optimisation, many-objective optimisation, self-adapting algorithms and formal analysis of fitness landscapes, highlighting research areas within the field of DMOO that should be addressed in future work.
{"title":"Challenges of Dynamic Multi-objective Optimisation","authors":"Mardé Helbig, A. Engelbrecht","doi":"10.1109/BRICS-CCI-CBIC.2013.49","DOIUrl":"https://doi.org/10.1109/BRICS-CCI-CBIC.2013.49","url":null,"abstract":"Dynamic multi-objective optimisation (DMOO) entails solving optimisation problems with more than one objective, where at least one objective changes over time. Normally at least two of the objectives are in conflict with one another. Therefore, a single solution does not exist and the goal of an algorithm is to find for each environment a set of solutions that are both diverse and as close as possible to the optimal trade-off solution set. Solving dynamic multi-objective optimisation problems (DMOOPs) is not a trivial task, since the field of DMOO has many challenges. This paper highlights these challenges, namely the selection of benchmark functions and performance measures, the analyses of obtained results and selecting a preferred solution from the set of trade-off solutions. In addition, this paper discusses emerging research areas within computational intelligence (CI), such as hyper-heuristics, constrained optimisation, many-objective optimisation, self-adapting algorithms and formal analysis of fitness landscapes, highlighting research areas within the field of DMOO that should be addressed in future work.","PeriodicalId":306195,"journal":{"name":"2013 BRICS Congress on Computational Intelligence and 11th Brazilian Congress on Computational Intelligence","volume":"46 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2013-09-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115820476","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2013-09-08DOI: 10.1109/BRICS-CCI-CBIC.2013.15
J. R. Bertini, Liang Zhao
Semi-supervised learning algorithms address the problem of learning from partially labeled data. However, most of the semi-supervised classification methods proposed in the literature considers a stationary distribution of data. Which means that future data patterns tend to conform to the data distribution presented in data set throughout the application lifetime. However, for plenty of new variety of applications, this expected scenario is not compatible to reality. Therefore, the research of semi-supervised methods which comprises nonstationary data classification is of a major concern nowadays. In this paper, the KAOGINCSSL algorithm, which copes with non-stationary semi-supervised learning, is analysed when using two different strategies to spread the labels to train the classifiers. The first consist of employing the inductive algorithm KAOGSS to build the classifier and the second relies on using the transductive algorithm PMTLA to spread the labels prior to build the classifier. Results regarding accuracy and processing time involving both algorithms when applied to non-stationary problems are presented.
{"title":"A Comparison of Two Purity-Based Algorithms When Applied to Semi-supervised Streaming Data Classification","authors":"J. R. Bertini, Liang Zhao","doi":"10.1109/BRICS-CCI-CBIC.2013.15","DOIUrl":"https://doi.org/10.1109/BRICS-CCI-CBIC.2013.15","url":null,"abstract":"Semi-supervised learning algorithms address the problem of learning from partially labeled data. However, most of the semi-supervised classification methods proposed in the literature considers a stationary distribution of data. Which means that future data patterns tend to conform to the data distribution presented in data set throughout the application lifetime. However, for plenty of new variety of applications, this expected scenario is not compatible to reality. Therefore, the research of semi-supervised methods which comprises nonstationary data classification is of a major concern nowadays. In this paper, the KAOGINCSSL algorithm, which copes with non-stationary semi-supervised learning, is analysed when using two different strategies to spread the labels to train the classifiers. The first consist of employing the inductive algorithm KAOGSS to build the classifier and the second relies on using the transductive algorithm PMTLA to spread the labels prior to build the classifier. Results regarding accuracy and processing time involving both algorithms when applied to non-stationary problems are presented.","PeriodicalId":306195,"journal":{"name":"2013 BRICS Congress on Computational Intelligence and 11th Brazilian Congress on Computational Intelligence","volume":"25 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2013-09-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116359370","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2013-09-08DOI: 10.1109/BRICS-CCI-CBIC.2013.56
A. Petrovsky
The paper describes the new tools for group sorting and ordering multi-attribute objects, when several versions of object may exist. For instance, values of attributes are estimated by several actors upon many quantitative and qualitative criteria or measured in different ways. These methods are based on the verbal decision analysis and the theory of multiset metric spaces. The developed techniques were applied to the multiple criteria selection of competitive R&D applications and the evaluation of project efficiency in the Russian Foundation for Basic Research.
{"title":"Multiset Tools for Group Multiple Criteria Decision Aiding","authors":"A. Petrovsky","doi":"10.1109/BRICS-CCI-CBIC.2013.56","DOIUrl":"https://doi.org/10.1109/BRICS-CCI-CBIC.2013.56","url":null,"abstract":"The paper describes the new tools for group sorting and ordering multi-attribute objects, when several versions of object may exist. For instance, values of attributes are estimated by several actors upon many quantitative and qualitative criteria or measured in different ways. These methods are based on the verbal decision analysis and the theory of multiset metric spaces. The developed techniques were applied to the multiple criteria selection of competitive R&D applications and the evaluation of project efficiency in the Russian Foundation for Basic Research.","PeriodicalId":306195,"journal":{"name":"2013 BRICS Congress on Computational Intelligence and 11th Brazilian Congress on Computational Intelligence","volume":"20 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2013-09-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116517323","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2013-09-08DOI: 10.1109/BRICS-CCI-CBIC.2013.68
Vladimiro Miranda, Rui Alves
This paper explores, with numerical case studies, the performance of an optimization algorithm that is a variant of EPSO, the Evolutionary Particle Swarm Optimization method. EPSO is already a hybrid approach that may be seen as a PSO with self-adaptive weights or an Evolutionary Programming approach with a self-adaptive recombination operator. The new hybrid DEEPSO retains the self-adaptive properties of EPSO but borrows the concept of rough gradient from Differential Evolution algorithms. The performance of DEEPSO is compared to a well-performing EPSO algorithm in the optimization of problems of the fixed cost type, showing consistently better results in the cases presented.
{"title":"Differential Evolutionary Particle Swarm Optimization (DEEPSO): A Successful Hybrid","authors":"Vladimiro Miranda, Rui Alves","doi":"10.1109/BRICS-CCI-CBIC.2013.68","DOIUrl":"https://doi.org/10.1109/BRICS-CCI-CBIC.2013.68","url":null,"abstract":"This paper explores, with numerical case studies, the performance of an optimization algorithm that is a variant of EPSO, the Evolutionary Particle Swarm Optimization method. EPSO is already a hybrid approach that may be seen as a PSO with self-adaptive weights or an Evolutionary Programming approach with a self-adaptive recombination operator. The new hybrid DEEPSO retains the self-adaptive properties of EPSO but borrows the concept of rough gradient from Differential Evolution algorithms. The performance of DEEPSO is compared to a well-performing EPSO algorithm in the optimization of problems of the fixed cost type, showing consistently better results in the cases presented.","PeriodicalId":306195,"journal":{"name":"2013 BRICS Congress on Computational Intelligence and 11th Brazilian Congress on Computational Intelligence","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2013-09-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122598639","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2013-09-08DOI: 10.1109/BRICS-CCI-CBIC.2013.75
Fabiano T. Novais, L. Batista, Agnaldo J. Rocha, F. Guimarães
In this paper, we propose a hybrid Multiobjective Estimation of Distribution Algorithm based on Artificial Bee Colonies and Clusters (MOEDABC) to solve multiobjective optimization problems with continuous variables. This algorithm is inspired in the organization and division of work in a bee colony and employs techniques from estimation of distribution algorithms. To improve some estimations we also employ clustering methods in the objective space. In the MOEDABC model, the colony consists of four groups of bees, each of which with its specific role in the colony: employer bees, onlookers, farmers and scouts. Each role is associated to specific tasks in the optimization process and employs different estimation of distribution methods. By combining estimation of distribution, clusterization of the objective domain, and the crowding distance assignment of NSGA-II, it was possible to extract more information about the optimization problem, thus enabling an efficient solution of large scale decision variable problems. Regarding the test problems, quality indicators, and GDE3, MOEA/D and NSGA-II methods, the combination of strategies incorporated into the MOEDABC algorithm has presented competitive results, which indicate this method as a useful optimization tool for the class of problems considered.
{"title":"A Multiobjective Estimation of Distribution Algorithm Based on Artificial Bee Colony","authors":"Fabiano T. Novais, L. Batista, Agnaldo J. Rocha, F. Guimarães","doi":"10.1109/BRICS-CCI-CBIC.2013.75","DOIUrl":"https://doi.org/10.1109/BRICS-CCI-CBIC.2013.75","url":null,"abstract":"In this paper, we propose a hybrid Multiobjective Estimation of Distribution Algorithm based on Artificial Bee Colonies and Clusters (MOEDABC) to solve multiobjective optimization problems with continuous variables. This algorithm is inspired in the organization and division of work in a bee colony and employs techniques from estimation of distribution algorithms. To improve some estimations we also employ clustering methods in the objective space. In the MOEDABC model, the colony consists of four groups of bees, each of which with its specific role in the colony: employer bees, onlookers, farmers and scouts. Each role is associated to specific tasks in the optimization process and employs different estimation of distribution methods. By combining estimation of distribution, clusterization of the objective domain, and the crowding distance assignment of NSGA-II, it was possible to extract more information about the optimization problem, thus enabling an efficient solution of large scale decision variable problems. Regarding the test problems, quality indicators, and GDE3, MOEA/D and NSGA-II methods, the combination of strategies incorporated into the MOEDABC algorithm has presented competitive results, which indicate this method as a useful optimization tool for the class of problems considered.","PeriodicalId":306195,"journal":{"name":"2013 BRICS Congress on Computational Intelligence and 11th Brazilian Congress on Computational Intelligence","volume":"191 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2013-09-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131684476","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2013-09-08DOI: 10.1109/BRICS-CCI-CBIC.2013.103
Isaac de L. Oliveira Filho, Otaciana G. R. Santiago, A. Canuto, B. Bedregal
In this paper, we propose a comparative analysis of the use of cryptography and transformation functions to be used as biometric (signature) template protection methods. The main goal is to investigate the increasement of the biometric dataset security as well as the performance of the protected dataset in the biometric-based systems. We use the well-elaborated structures for pattern recognition (ensembles systems) on unprotected and protected dataset to measure the performance of the biometric template protection methods used in this research. The results would allow us to define the most secure used protection method which keeps an acceptable accuracy level at the same time.
{"title":"Analyzing Ensemble Systems for Protected Biometric Data","authors":"Isaac de L. Oliveira Filho, Otaciana G. R. Santiago, A. Canuto, B. Bedregal","doi":"10.1109/BRICS-CCI-CBIC.2013.103","DOIUrl":"https://doi.org/10.1109/BRICS-CCI-CBIC.2013.103","url":null,"abstract":"In this paper, we propose a comparative analysis of the use of cryptography and transformation functions to be used as biometric (signature) template protection methods. The main goal is to investigate the increasement of the biometric dataset security as well as the performance of the protected dataset in the biometric-based systems. We use the well-elaborated structures for pattern recognition (ensembles systems) on unprotected and protected dataset to measure the performance of the biometric template protection methods used in this research. The results would allow us to define the most secure used protection method which keeps an acceptable accuracy level at the same time.","PeriodicalId":306195,"journal":{"name":"2013 BRICS Congress on Computational Intelligence and 11th Brazilian Congress on Computational Intelligence","volume":"79 1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2013-09-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132175355","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2013-09-08DOI: 10.1109/BRICS-CCI-CBIC.2013.91
Thierry Martinez, L. Vitorino, F. Fages, A. Aggoun
Bin packing is a classical combinatorial optimization problem which has a wide range of real-world applications in industry, logistics, transport, parallel computing, circuit design and other domains. While usually presented as discrete problems, we consider here continuous packing problems including curve shapes, and model these problems as continuous optimization problems with a multi-objective function combining non-overlapping with minimum bin size constraints. More specifically, we consider the covariance matrix adaptation evolution strategy (CMA-ES) with a non-overlapping and minimum size objective function in either two or three dimensions. Instead of taking the intersection area as measure of overlap, we propose other measures, monotonic with respect to the intersection area, to better guide the search. In order to compare this approach to previous work on bin packing, we first evaluate CMA-ES on Korf's benchmark of consecutive sizes square packing problems, for which optimal solutions are known, and on a benchmark of circle packing problems. We show that on square packing, CMA-ES computes solutions at typically 14% of the optimal cost, with the time limit given by the best dedicated algorithm for computing optimal solutions, and that on circle packing, the computed solutions are at 2% of the best known solutions. We then consider generalizations of this benchmark to mixed squares and circles, boxes, spheres and cylinders packing problems, and study a real-world problem for loading boxes and cylinders in containers. These hard problems illustrate the interesting trade-off between generality and efficiency in this approach.
{"title":"On Solving Mixed Shapes Packing Problems by Continuous Optimization with the CMA Evolution Strategy","authors":"Thierry Martinez, L. Vitorino, F. Fages, A. Aggoun","doi":"10.1109/BRICS-CCI-CBIC.2013.91","DOIUrl":"https://doi.org/10.1109/BRICS-CCI-CBIC.2013.91","url":null,"abstract":"Bin packing is a classical combinatorial optimization problem which has a wide range of real-world applications in industry, logistics, transport, parallel computing, circuit design and other domains. While usually presented as discrete problems, we consider here continuous packing problems including curve shapes, and model these problems as continuous optimization problems with a multi-objective function combining non-overlapping with minimum bin size constraints. More specifically, we consider the covariance matrix adaptation evolution strategy (CMA-ES) with a non-overlapping and minimum size objective function in either two or three dimensions. Instead of taking the intersection area as measure of overlap, we propose other measures, monotonic with respect to the intersection area, to better guide the search. In order to compare this approach to previous work on bin packing, we first evaluate CMA-ES on Korf's benchmark of consecutive sizes square packing problems, for which optimal solutions are known, and on a benchmark of circle packing problems. We show that on square packing, CMA-ES computes solutions at typically 14% of the optimal cost, with the time limit given by the best dedicated algorithm for computing optimal solutions, and that on circle packing, the computed solutions are at 2% of the best known solutions. We then consider generalizations of this benchmark to mixed squares and circles, boxes, spheres and cylinders packing problems, and study a real-world problem for loading boxes and cylinders in containers. These hard problems illustrate the interesting trade-off between generality and efficiency in this approach.","PeriodicalId":306195,"journal":{"name":"2013 BRICS Congress on Computational Intelligence and 11th Brazilian Congress on Computational Intelligence","volume":"20 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2013-09-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128649898","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：