Search

Accurately predicting the performance of different optimization algorithms for previously unseen problem instances is crucial for high-performing algorithm selection and configuration techniques. In the context of numerical optimization, supervised regression approaches built on top of exploratory landscape analysis are becoming very popular. From the point of view of Machine Learning (ML), however, the approaches are often rather naive, using default regression or classification techniques without proper investigation of the suitability of the ML tools. With this work, we bring to the attention of our community the possibility to personalize regression models to specific types of optimization problems. Instead of aiming for a single model that works well across a whole set of possibly diverse problems, our personalized regression approach acknowledges that different models may suite different types of problems. Going one step further, we also investigate the impact of selecting not a single regression model per problem, but personalized ensembles. We test our approach on predicting the performance of numerical optimization heuristics on the BBOB benchmark collection., To appear in the Proceedings of Genetic and Evolutionary Computation Conference (GECCO 2021), ACM

The performance assessment of multi-objective optimization algorithms involves a user-preference-based selection of a single quality indicator used as a performance measure. A single quality indicator maps the approximation set (i.e., high-dimensional data) into a real value (i.e, one-dimensional data). However, it is well known that the selection of the quality indicator can have a huge impact on the benchmarking conclusions. This invites researchers to present only results for quality indicators that are in favor of the desired algorithm, or performing bias performance assessment. To go beyond this, we proposed a novel ranking scheme that reduces the bias in the user-preference selection by comparing the high-dimensional data of approximations sets and consequently provides more robust statistical results. The selection of a quality indicator is only required in cases when high-dimensional distributions of the approximation sets differ. By performing such analyses, experimental results show that the cases affected by the user-preference selection are reduced.

When designing a benchmark problem set, it is important to create a set of benchmark problems that are a good generalization of the set of all possible problems. One possible way of easing this difficult task is by using artificially generated problems. In this paper, one such single-objective continuous problem generation approach is analyzed and compared with the COCO benchmark problem set, a well know problem set for benchmarking numerical optimization algorithms. Using Exploratory Landscape Analysis and Singular Value Decomposition, we show that such representations allow us to further explore the relations between the problems by applying visualization and correlation analysis techniques, with the goal of decreasing the bias in benchmark problem assessment., To appear in the Proceedings of Genetic and Evolutionary Computation Conference Companion (GECCO 2021), ACM

The performance assessment of multi-objective optimization algorithms is a crucial task for investigating their behaviour. However, the selected quality indicators and statistical techniques used in comparison studies can have huge impact on the study results. A quality indicator transforms high-dimensional data (an approximation set) into one-dimensional data (a quality indicator), followed by a potential loss of high-dimensional information concerning the transformation. Comparison approaches typically involve a single quality indicator or an ensemble of quality indicators to address more quality criteria, which are predefined by the user. To provide more robust benchmarking for multi-objective optimization, we extended the DSCTool with three approaches that are ensembles of quality indicators and one novel approach that compare the high-dimensional distributions of the approximation sets and reduces the users' preference in the selection of quality indicators. The approaches are provided as web services for robust ranking and hypothesis testing, including a proper selection of an omnibus statistical test and post-hoc tests if needed.

Exploratory Landscape Analysis is a methodology for transforming the samples of an optimization problem into numerical descriptors called landscape features. Since Exploratory Landscape Analysis is sample based, recent studies have shown that the choice of the method used to collect the problem samples can have an effect on the calculation of landscape features.In our recent work, we have shown that certain landscape features are not invariant to even simple function transformations such as shifting or scaling. However, the analysis in our previous work was conducted using only Latin Hypercube Sampling. Since we are now aware that the choice of sampling method can affect the calculation of landscape features, this paper expands upon our earlier work by using a variety of different sampling methods. We show that different sampling methods do indeed have an effect on the invariance of landscape features, and present a list of landscape features that are invariant under all of our chosen sampling methods.

Solving hard optimization problems is one of the most important research topics due to the countless applications in different areas. Since solving such problems is of great importance, numerous metaheuristics were developed, many of which belong to the group of swarm intelligence optimization algorithms. In recent decades, there has been an explosion in the number of the proposed swarm intelligence algorithms most commonly compared to other metaheuristics using one statistic such as average or median which can lead to putting algorithms in different rankings even though there are only small differences between them. In order to provide more insights into swarm intelligence algorithms’ performance, a deep statistical comparison is used. Five representative swarm intelligence optimization algorithms are ranked based on the obtained solutions values and their distribution in the search space while solving the CEC2013 benchmark functions. The used analysis differentiates algorithms that have statistically significant performance and measure the qualities of the exploration and the exploitation abilities of the tested algorithms.

Automated per-instance algorithm selection and configuration have shown promising performances for a number of classic optimization problems, including satisfiability, AI planning, and TSP. The techniques often rely on a set of features that measure some characteristics of the problem instance at hand. In the context of black-box optimization, these features have to be derived from a set of $(x,f(x))$ samples. A number of different features have been proposed in the literature, measuring, for example, the modality, the separability, or the ruggedness of the instance at hand. Several of the commonly used features, however, are highly correlated. While state-of-the-art machine learning techniques can routinely filter such correlations, they hinder explainability of the derived algorithm design techniques. We therefore propose in this work to pre-process the measured (raw) landscape features through representation learning. More precisely, we show that a linear dimensionality reduction via matrix factorization significantly contributes towards a better detection of correlation between different problem instances -- a key prerequisite for successful automated algorithm design., To appear in Proc. of 2020 IEEE Symposium Series on Computational Intelligence (SSCI 2020)

When making statistical analysis of single-objective optimization algorithms&rsquo, performance, researchers usually estimate it according to the obtained optimization results in the form of minimal/maximal values. Though this is a good indicator about the performance of the algorithm, it does not provide any information about the reasons why it happens. One possibility to get additional information about the performance of the algorithms is to study their exploration and exploitation abilities. In this paper, we present an easy-to-use step by step pipeline that can be used for performing exploration and exploitation analysis of single-objective optimization algorithms. The pipeline is based on a web-service-based e-Learning tool called DSCTool, which can be used for making statistical analysis not only with regard to the obtained solution values but also with regard to the distribution of the solutions in the search space. Its usage does not require any special statistic knowledge from the user. The gained knowledge from such analysis can be used to better understand algorithm&rsquo, s performance when compared to other algorithms or while performing hyperparameter tuning.

The performance measures and statistical techniques selected affect the conclusions we can draw on the behavior of the algorithms. For this reason, we propose more robust performance statistics for addressing statistical and practical significance, as well as investigating the exploration and exploitation powers of stochastic optimization algorithms. They are introduced by the Deep Statistical Comparison approach and its variants. Its implementations are available as part of the DSCTool, which provides web services for robust ranking and hypothesis testing, including a proper selection of an omnibus statistical test and post-hoc tests if needed.

To transfer the learned knowledge that is coming from benchmarking studies, which involve stochastic optimization algorithms, we should find a way to decide if the statistical significance between their performance is also important for real-world applications. For this reason, we have recently proposed a practical Deep Statistical Comparison (pDSC). It takes into account practical significance when making a statistical comparison of meta-heuristic stochastic optimization algorithms for single-objective optimization problems. Experimental results showed that our recently proposed approach provided very promising results.

When working on a new stochastic optimization algorithm, one task that should be performed is to compare its performance with those of state-of-the-art algorithms. The literature suggests that the most commonly applied approaches for comparing algorithms' performances use statistical analyses. However, to provide a more meaningful explanation about algorithms' performances, we propose a methodology, named performance2vec, which computes a vector representation of each algorithm's performance by embedding it in some performance space determined by a set of benchmark problems. Experimental results show that the proposed embeddings, paired with clustering approaches, provide a more in-depth explanation regarding the algorithms' performance by exploring the relations between them and the benchmark problems.

To visually present the overall performance of several algorithms tested on several benchmark problems on one plot, we present a machine learning approach, called performViz. It allows one to clearly see, from a single plot, which algorithms are most suited for a given problem, the influence of each problem on the overall algorithm performance and similarities among both algorithms and problems. It consists of four steps: i) selecting a performance measure, ii) selecting a statistic to calculate a representative value of the performance measure for each algorithm on each problem, iii) performing hierarchical clustering, and iv) using heatmaps to visualize the clustering result.

Selecting the relevant algorithm for a given problem is a crucial first step for achieving good optimization algorithm performance. Exploratory Landscape Analysis can help with this problem by calculating landscape features that numerically describe individual problems. To understand the problem space in single-objective numerical optimization, we recently presented a preliminary study of how Exploratory Landscape Analysis can be used to visualize different optimization benchmark problems, with the ultimate goal of visualizing problems that are similar to one another close together. In this paper, we examine how the selection of landscape features affects such a visualization, and show that proper preprocessing of landscape features is crucial for producing a good visualization. In particular, we show that only a subset of landscape features is invariant to simple transforms such as translation and scaling. Further, we examine how such an approach can be used to visually compare problems from different optimization benchmark sets.

International audience

By performing data analysis, statistical approaches are highly welcome to explore the data. Nowadays with the increases in computational power and the availability of big data in different domains, it is not enough to perform exploratory data analysis (descriptive statistics) to obtain some prior insights from the data, but it is a requirement to apply higher-level statistics that also require much greater knowledge from the user to properly apply them. One research area where proper usage of statistics is important is multi-objective optimization, where the performance of a newly developed algorithm should be compared with the performances of state-of-the-art algorithms. In multi-objective optimization, we are dealing with two or more usually conflicting objectives, which result in high dimensional data that needs to be analyzed. In this paper, we present a web-service-based e-Learning tool called DSCTool that can be used for performing a proper statistical analysis for multi-objective optimization. The tool does not require any special statistics knowledge from the user. Its usage and the influence of a proper statistical analysis is shown using data taken from a benchmarking study performed at the 2018 IEEE CEC (The IEEE Congress on Evolutionary Computation) is appropriate. Competition on Evolutionary Many-Objective Optimization.

International audience

This paper addresses the problem of missing data in food composition databases (FCDBs). The missing data can be either for selected foods or for specific components only. Most often, the problem is solved by human experts subjectively borrowing data from other FCDBs, for data estimation or imputation. Such an approach is not only time-consuming but may also lead to wrong decisions as the value of certain components in certain foods may vary from database to database due to differences in analytical methods. To ease missing-data borrowing and increase the quality of missing-data selection, we propose a new computer-based methodology, named MIGHT - Missing Nutrient Value Imputation UsinG Null Hypothesis Testing, that enables optimal selection of missing data from different FCDBs. The evaluation on a subset of European FCDBs, available through EuroFIR and complied with the Food data structure and format standard BS EN 16104 published in 2012, proves that, in more than 80% of selected cases, MIGHT gives more accurate results than techniques currently applied for missing value imputation in FCDBs. MIGHT deals with missing data in FCDBs by introducing rules for missing data imputation based on the idea that proper statistical analysis can decrease the error of data borrowing.

Understanding of exploration and exploitation powers of meta-heuristic stochastic optimization algorithms is very important for algorithm developers. For this reason, we have recently proposed an approach for making a statistical comparison of meta-heuristic stochastic optimization algorithms according to the distribution of the solutions in the search space, which is also presented in this paper. Its main contribution is the support to identify exploration and exploitation powers of the compared algorithms. This is especially important when dealing with multimodal search spaces, which consist of many local optima with similar values, and large-scale continuous optimization problems, where it is hard to understand the reasons for the differences in performances. Experimental results showed that our recently proposed approach gives very promising results.

The GECCO Workshop on Real-Parameter Black-Box Optimization Benchmarking Series is a series of benchmarking workshops held every year since 2009 that evaluates the performance of new optimization algorithms. Originally, the workshop organizers provided results for every year the workshop took place. In this article, we directly compare algorithms from every workshop held from 2009 to 2018. We analyze the compared algorithms on the noiseless benchmark function set using the competition's official empirical approach, and with a statistical approach called Deep Statistical Comparison. Our goal is to investigate how algorithm performance has evolved throughout the years, and to show differences between empirical and statistical approaches to evaluating results.

The Special Sessions and Competitions on Real-Parameter Single Objective Optimization are benchmarking competitions held every year since 2013 that are used to evaluate the performance of new optimization algorithms.One flaw of these competitions is that algorithms are compared only to other algorithms submitted in the same year, not with algorithms submitted in previous years of the competition, so it can make comparison between all algorithms troublesome. Almost every year uses different benchmark functions, so the results between the years are not directly comparable. As a result, the winner of the most recent competition might not necessarily be significantly better than the winners of previous years.In this article, we directly compare winners of every competition held from 2013 to 2018 and present the results of this comparison. We use a benchmark set that consists of all test functions used by the competition throughout these years. We compare them on benchmark functions grouped by dimension (10, 30, 50, 100) and by year (2013, 2014, 2015, 2017). This allows us on one hand to see which algorithms perform best at specific dimensions, while grouping by year shows effects of parameter tuning on the end results.We present the results of these comparisons and find that later competition winners are not statistically better than algorithms from previous years in a general sense on every problem and dimensionality.

The multi-objective optimization approach has a large influence in the industrial production scheduling. The goal of such optimization is to find a production schedule that satisfies different, usually contradictory, production and business constraints. In the paper, memetic versions of three multi-objective algorithms with different approaches to problem solving are implemented. The customized reproduction operators and local search procedures are also used. These memetic algorithms are applied to real order-lists from a production company. It is shown that the multi-objective approaches are able to find high-quality solutions, also when quick respond is required to adapt to dynamic business conditions. According to the results it is concluded that for the two tested real-world problems the IBEA confirmed its superiority over the NSGA-II and SPEA2.

Making a statistical comparison of meta-heuristic multi-objective optimization algorithms is crucial for identifying the strengths and weaknesses of a newly proposed algorithm. Currently, state-of-the-art comparison approaches involve user-preference-based selection of a single quality indicator or an ensemble of quality indicators as a comparison metric. Using these quality indicators, high-dimensional data is transformed into one-dimensional data. By doing this, information contained in the high-dimensional space can be lost, which will affect the results of the comparison. To avoid losing this information, we propose a novel ranking scheme that compares the distributions of high-dimensional data. Experimental results show that the proposed approach reduces potential information loss when statistical significance is not observed in high-dimensional data. Consequently, the selection of a quality indicator is required only in cases when statistical significance is observed in high-dimensional data. With this the cases that are affected by the user preference selection are reduced.

DSCTool is a statistical tool for comparing performance of stochastic optimization algorithms on a single benchmark function (i.e. single-problem analysis) or a set of benchmark functions (i.e., multiple-problem analysis). DSCTool implements a recently proposed approach, called Deep Statistical Comparison (DSC), and its variants. DSC ranks optimization algorithms by comparing distributions of obtained solutions for a problem instead of using a simple descriptive statistic such as the mean or the median. The rankings obtained for an individual problem give the relations between the performance of the applied algorithms. To compare optimization algorithms in the multiple-problem scenario, an appropriate statistical test must be applied to the rankings obtained for a set of problems. The main advantage of DSCTool are its REST web services, which means all its functionalities can be accessed from any programming language. In this paper, we present the DSCTool in detail with examples for its usage.

In this paper, we propose an extension of a recently proposed Deep Statistical Comparison (DSC) approach, called practical Deep Statistical Comparison (pDSC), which takes into account practical significance when making a statistical comparison of meta-heuristic stochastic optimization algorithms for single-objective optimization. For achieving practical significance, two variants of the standard DSC ranking scheme are proposed. The first is called sequential pDSC, and takes into account practical significance by preprocessing of the independent optimization runs in a sequential order. The second is called Monte Carlo pDSC, and avoids any dependency of practical significance with regard to the ordering of optimization runs. The analysis of identifying practical significance on benchmark tests for single-objective problems, shows that for some cases, both variants of pDSC compared to the Chess Rating System for Evolutionary Algorithms (CRS4EAs) approach give different conclusions. Preprocessing for practical significance is carried out in a similar way, but there are cases when the conclusion for practical significance differ, which comes from the different statistical concepts used to identify practical significance.

In this paper a recently proposed approach for making a statistical comparison of meta-heuristic stochastic optimization algorithms is presented. The main contribution of this approach is that the ranking scheme is based on the whole distribution, instead of using only one statistic to describe the distribution, such as average or median. Experimental results showed that our approach gives more robust results compared to state-of-the-art approaches in case when the results are affected by outliers or by statistical insignificant differences that could exist between data values.

This paper identifies the requirements for computer-supported food matching, in order to address not only national and European but also international current related needs and represents an integrated research contribution of the FP7 EuroDISH project. The available classification and coding systems and the specific problems of food matching are summarized and a new concept for food matching based on optimization methods and machine-based learning is proposed. To illustrate and test this concept, a study has been conducted in four European countries (i.e., Germany, The Netherlands, Italy and the UK) using different classification and coding systems. This real case study enabled us to evaluate the new food matching concept and provide further recommendations for future work. In the first stage of the study, we prepared subsets of food consumption data described and classified using different systems, that had already been manually matched with national food composition data. Once the food matching algorithm was trained using this data, testing was performed on another subset of food consumption data. Experts from different countries validated food matching between consumption and composition data by selecting best matches from the options given by the matching algorithm without seeing the result of the previously made manual match. The evaluation of study results stressed the importance of the role and quality of the food composition database as compared to the selected classification and/or coding systems and the need to continue compiling national food composition data as eating habits and national dishes still vary between countries. Although some countries managed to collect extensive sets of food consumption data, these cannot be easily matched with food composition data if either food consumption or food composition data are not properly classified and described using any classification and coding systems. The study also showed that the level of human expertise played an important role, at least in the training stage. Both sets of data require continuous development to improve their quality in dietary assessment.

To find the strengths and weaknesses of a new multi-objective optimization algorithm, we need to compare its performance with the performances of the state-of-the-art algorithms. Such a comparison involves a selection of a performance metric, a set of benchmark problems, and a statistical test to ensure that the results are statistical significant. There are also studies in which instead of using one performance metric, a comparison is made using a set of performance metrics. All these studies assume that all involved performance metrics are equal. In this paper, we introduce a data-driven preference-based approach that is a combination of multiple criteria decision analysis with deep statistical rankings. The approach ranks the algorithms for each benchmark problem using the preference (the influence) of each performance metric that is estimated using its entropy. Experimental results show that this approach achieved similar rankings to a previously proposed method, which is based on the idea of the majority vote, where all performance metrics are assumed equal. However, as it will be shown, this approach can give different rankings because it is based not only on the idea of counting wins, but also includes information about the influence of each performance metric.

Benchmarking theory in evolutionary computation research is a crucial task that should be properly applied in order to evaluate the performance of a newly introduced evolutionary algorithm with performance of state-of-the-art algorithms. Benchmarking theory is related to three main questions: which problems to choose, how to setup experiments, and how to evaluate performance. In this paper, we evaluate the impact of different already established statistical ranking schemes that can be used for evaluation of performance in benchmarking practice for evolutionary computation. Experimental results obtained on Black-Box Benchmarking 2015 showed that different statistical ranking schemes, used on the same benchmarking data, can lead to different benchmarking results. For this reason, we examined the merits and issues of each of them regarding benchmarking practices.

In this paper, a study of how to compare the performance of multi-objective stochastic optimization algorithms using quality indicators and Deep Statistical Comparison (DSC) approach is presented. DSC is a recently proposed approach for statistical comparison of meta-heuristic stochastic optimization algorithms over single-objective problems. The main contribution of DSC is the ranking scheme that is based on the whole distribution, instead of using only one statistic such as average or median. Experimental results performed by using 6 multi-objective stochastic optimization algorithms on 16 test problems show that the DSC gives more robust results compared to some standard statistical approaches that are recommended for a comparison of multi-objective stochastic optimization algorithms according to some quality indicator.

The European Food Safety Authority has developed a standardized food classification and description system called FoodEx2. It uses facets to describe food properties and aspects from various perspectives, making it easier to compare food consumption data from different sources and perform more detailed data analyses. However, both food composition data and food consumption data, which need to be linked, are lacking in FoodEx2 because the process of classification and description has to be manually performed—a process that is laborious and requires good knowledge of the system and also good knowledge of food (composition, processing, marketing, etc.). In this paper, we introduce a semi-automatic system for classifying and describing foods according to FoodEx2, which consists of three parts. The first involves a machine learning approach and classifies foods into four FoodEx2 categories, with two for single foods: raw (r) and derivatives (d), and two for composite foods: simple (s) and aggregated (c). The second uses a natural language processing approach and probability theory to describe foods. The third combines the result from the first and the second part by defining post-processing rules in order to improve the result for the classification part. We tested the system using a set of food items (from Slovenia) manually-coded according to FoodEx2. The new semi-automatic system obtained an accuracy of 89% for the classification part and 79% for the description part, or an overall result of 79% for the whole system.

This paper presents a study on making a statistical comparison of multi-objective optimization algorithms using an ensemble of quality indicators together with a deep statistical comparison (DSC) approach. The DSC approach has been recently proposed for statistically comparing meta-heuristic stochastic optimization algorithms for single-objective problems. The DSC ranking scheme is based on the whole distribution, rather than on one statistic such as either the average or the median. This study uses two ensemble combiners to rank and compare algorithms using the DSC ranking scheme for each quality indicator for a given problem. Experimental results performed using 3 multi-objective optimization algorithms on 16 test problems show that ensembles of quality indicators with transformed DSC rankings give more robust results than when the same ensembles are used with transformed rankings obtained by some standard ranking schemes.

Evidence-based dietary information represented as unstructured text is a crucial information that needs to be accessed in order to help dietitians follow the new knowledge arrives daily with newly published scientific reports. Different named-entity recognition (NER) methods have been introduced previously to extract useful information from the biomedical literature. They are focused on, for example extracting gene mentions, proteins mentions, relationships between genes and proteins, chemical concepts and relationships between drugs and diseases. In this paper, we present a novel NER method, called drNER, for knowledge extraction of evidence-based dietary information. To the best of our knowledge this is the first attempt at extracting dietary concepts. DrNER is a rule-based NER that consists of two phases. The first one involves the detection and determination of the entities mention, and the second one involves the selection and extraction of the entities. We evaluate the method by using text corpora from heterogeneous sources, including text from several scientifically validated web sites and text from scientific publications. Evaluation of the method showed that drNER gives good results and can be used for knowledge extraction of evidence-based dietary recommendations.

Geographical information systems are ideal candidates for the application of parallel programming techniques, mainly because they usually handle large data sets. To help us deal with complex calculations over such data sets, we investigated the performance constraints of a classic master-worker parallel paradigm over a message-passing communication model. To this end, we present a new approach that employs an external database in order to improve the calculation/communication overlap, thus reducing the idle times for the worker processes. The presented approach is implemented as part of a parallel radio-coverage prediction tool for the GRASS environment. The prediction calculation employs digital elevation models and land-usage data in order to analyze the radio coverage of a geographical area. We provide an extended analysis of the experimental results, which are based on real data from an LTE network currently deployed in Slovenia. Based on the results of the experiments, which were performed on a computer cluster, the new approach exhibits better scalability than the traditional master-worker approach. We successfully tackled real-world data sets, while greatly reducing the processing time and saturating the hardware utilization., 13 pages, 12 figures and 2 tables. International Journal of Geographical Information Science, 2014

Many promising optimization algorithms for solving numerical optimization problems come from population-based metaheuristics. A few of them are based on Swarm-Intelligence Algorithms, which are inspired by the collective behavior of social organisms. One of the most successful of such algorithms is the Differential Ant-Stigmergy Algorithm (DASA), which uses stigmergy, a method of communication in emergent systems where the individual parts (artificial ants) of the system communicate with one another by modifying their local environment (pheromone intensity). The main characteristic of the DASA is its underlying structure (pheromone graph) that uses discrete steps to move through a continuous search space. As a consequence of this the search-space movement is in some way limited and the algorithm's time/space complexity is increased. In order to overcome the problem an improved algorithm called the Continuous Differential Ant-Stigmergy Algorithm (CDASA) is proposed and then benchmarked on standard benchmark functions. This benchmarking showed that the CDASA performs better than the DASA, especially at lower dimensions, that its time/space complexity is decreased, and that the algorithm code is simplified. As such, the CDASA is more suitable for parallel implementations on General-Purpose Graphic Processing Units. Compared to the Swarm-Intelligence Algorithms presented in this paper, the CDASA is the best-performing algorithm and competitive to the state-of-the-art algorithms belonging to different metaheuristic approaches.

Arithmetic coding is a lossless compression algorithm with variable-length source coding. It is more flexible and efficient than the well-known Huffman coding. In this paper we present a non-adaptive FPGA implementation of a multi-alphabet arithmetic coding with separated statistical model of the data source. The alphabet of the data source is a 256-symbol ASCII character set and does not include the special end-of-file symbol. No context switching is used in the proposed design which gives maximal throughput without pipelining. We have synthesized the design for Xilinx FPGA devices and used their built-in hardware resources. K e y w o r d s: arithmetic coding, compression algorithm

Planning problems can be solved with a large variety of different approaches, and a significant amount of work has been devoted to the automation of planning processes using different kinds of methods. This paper focuses on the use of specific local search algorithms for real-world production planning based on experiments with real-world data, and presents an adapted local search guided by evolutionary metaheuristics. To make algorithms efficient, many specifics need to be considered and included in the problem solving. We demonstrate that the use of specialized local searches can significantly improve the convergence and efficiency of the algorithm. The paper also includes an experimental study of the efficiency of two memetic algorithms, and presents a real-world software implementation for the production planning.

The paper presents the evaluation of memetic algorithm with the real production planning problem, which consists of many constraints. Additionally, there are some orders with fixed deadlines that have to be produced on the exact day. The memetic algorithm with specialized local searches is able to effectively improve solutions acquired by genetic algorithm even if many orders with fixed deadlines are given.

PurposeThe purpose of this paper is to present an algorithm for global optimization of high‐dimensional real‐parameter cost functions.Design/methodology/approachThis optimization algorithm, called differential ant‐stigmergy algorithm (DASA), based on a stigmergy observed in colonies of real ants. Stigmergy is a method of communication in decentralized systems in which the individual parts of the system communicate with one another by modifying their local environment.FindingsThe DASA outperformed the included differential evolution type algorithm in convergence on all test functions and also obtained better solutions on some test functions.Practical implicationsThe DASA may find applications in challenging real‐life optimization problems such as maximizing the empirical area under the receiver operating characteristic curve of glycomics mass spectrometry data and minimizing the logistic leave‐one‐out calculation measure for the gene‐selection criterion.Originality/valueThe DASA is one of the first ant‐colony optimization‐based algorithms proposed for global optimization of the high‐dimensional real‐parameter problems.

This article presents an optimization method used at the electric-motor design. The goal of the optimization was to find the geometrical parameter values that would generate the rotor and the stator geometries with minimum power losses. A new, distributed version of the multilevel ant-stigmergy algorithm (MASA) was proposed to solve this optimization problem. The roots of the MASA can be found in the ant-colony optimization metaheuristic. The usefulness and efficiency of some sequential population-based algorithms (electromagnetism-like, particle swarm, evolutionary, differential evolution) as well as a sequential version of the MASA for solving the problem of minimizing the losses in an electric motor have recently been reported in the literature. Results showed that the MASA outperformed all the other algorithms. Due to a time-consuming solution evaluation, however, all the sequential algorithms were inefficient in terms of time. In order to remedy this shortcoming, a new, efficient distributed implementation of the MASA is presented. In addition, we have shown that with distributed computing the computation time can be drastically reduced (from one day to a few hours) without any noticeable reduction in the quality of the solution.

The efficiency of universal electric motors that are widely used in home appliances can be improved by optimizing the geometry of the rotor and the stator. Expert designers traditionally approach this task by iteratively evaluating candidate designs and improving them according to their experience. However, the existence of reliable numerical simulators and powerful stochastic optimization techniques make it possible to automate the design procedure. We present a comparative study of six stochastic optimization algorithms in designing optimal rotor and stator geometries of a universal electric motor where the primary objective is to minimize the motor power losses. We compare three methods from the domain of evolutionary computation, generational evolutionary algorithm, steady-state evolutionary algorithm and differential evolution, two particle-based methods, particle-swarm optimization and electromagnetism-like algorithm, and a recently proposed multilevel ant stigmergy algorithm. By comparing their performance, the most efficient method for solving the problem is identified and an explanation of its success is offered.

The Multilevel Ant Stigmergy Algorithm (MASA) is a new approach to solving multi-parameter problems based on stigmergy, a type of collective work that can be observed in nature. In this paper we evaluate the performance of MASA regarding its applicability as numerical optimization techniques. The evaluation is performed with several widely used benchmarks functions, as well as on an industrial case study. We also compare the MASA with Differential Evolution, well-known numerical optimization algorithm. The average solution obtained with the MASA was better than a solution recently found using Differential Evolution.

Many real-world engineering problems can be expressed in terms of partial differential equations and solved by using the finite-element method, which is usually parallelised, i.e. the mesh is divided among several processors. To achieve high parallel efficiency it is important that the mesh is partitioned in such a way that workloads are well balanced and interprocessor communication is minimised. In this paper we present an enhancement of a technique that uses a nature-inspired metaheuristic approach to achieve higher-quality partitions. The so-called multilevel ant-colony algorithm, which is a relatively new metaheuristic search technique for solving optimisation problems, was applied and studied, and the possible parallelisation of this algorithm is discussed. The multilevel ant-colony algorithm performed very well and is superior to classical k-METIS and Chaco algorithms; it is even comparable with the combined evolutionary/multilevel scheme used in the JOSTLE evolutionary algorithm and returned solutions that are better than the currently available solutions in the Graph Partitioning Archive.

An initiative was introduced in one of the production facilities of Germany's E.G.O. Group in order to enhance its SAP information system with a custom-made application for production-scheduling optimization. The goal of the optimization is to find a production schedule that satisfies different, contradictory production and business constraints. We show the challenges faced in the application of the multi-objective optimization approach, which is gaining influence in the management of production scheduling. We implement a memetic version of the Indicator-Based Evolutionary Algorithm with customized reproduction operators and local search procedures to find a set of feasible, non-dominated solutions. Such a memetic algorithm was applied to two real order lists from the production company. Additionally, we also lay out an efficient presentation of the multi-objective results for an expert's support in decision making. This provides the management with the possibility to gain additional insights into how the production schedule dynamically reacts to changes in the decision criteria. We show that the multi-objective approach is able to find high-quality solutions, which enables flexibility when it comes to quickly adapting to specific business conditions.