Evolutionary Algorithm

The p-hub median problem (PHMP) is a fundamental NP-hard network design problem with wide applications in logistics, transportation, and telecommunication systems. Hyperheuristics have demonstrated competitive performance for large-scale hub location instances by dynamically selecting lowlevel heuristics during the search process. However, existing approaches lack explicit mechanisms to regulate heuristic diversity, often leading to premature convergence and structural stagnation. This paper proposes an Entropy-Constrained Hyperheuristic (ECHH) that integrates an information-theoretic diversity control mechanism into heuristic selection. By enforcing a lower bound on the Shannon entropy of the heuristic usage distribution, the proposed framework transforms diversification into a measurable and controllable property. The entropy threshold parameter governs the exploration-exploitation balance and induces a structural phase transition in search behavior. The method is evaluated on the Australian Post 200-node benchmark dataset under multiple hub densities and entropy levels. Experimental results show that moderate entropy thresholds significantly improve robustness, basin exploration, and average solution quality compared to unconstrained selection, while excessively high entropy reduces convergence speed. The findings demonstrate that structural diversity control acts as a regularization mechanism in hyperheuristic search and provides a principled approach to balancing exploration and exploitation in large-scale combinatorial optimization.

Groups at MIT and NYU have collected a dataset of millions of tiny colour images from the web. It is, in principle, an excellent dataset for unsupervised training of deep generative models, but previous researchers who have tried this have found it dicult to learn a good set of lters from the images. We show how to train a multi-layer generative model that learns to extract meaningful features which resemble those found in the human visual cortex. Using a novel parallelization algorithm to distribute the work among multiple machines connected on a network, we show how training such a model can be done in reasonable time. A second problematic aspect of the tiny images dataset is that there are no reliable class labels which makes it hard to use for object recognition experiments. We created two sets of reliable labels. The CIFAR-10 set has 6000 examples of each of 10 classes and the CIFAR-100 set has 600 examples of each of 100 non-overlapping classes. Using these labels, we show that object recognition is signicantly improved by pre-training a layer of features on a large set of unlabeled tiny images. C Labeler instruction sheet 52 D CIFAR-100 class structure 54

Most evolutionary algorithms optimize the information from good solutions found in the population. A selection method discards the below-average solutions assuming that they do not contribute any information to update the probabilistic models. This work develops an algorithm called Coincidence algorithm (COIN) which merges negative correlation learning into the optimization process. A knight's tour problem, one of NP-hard multimodal Hamiltonian path problems, is tested with COIN. The results show that COIN is a competitive algorithm in converging to better solutions and maintaining diverse solutions to solve combinatorial optimization problems.

The proliferation of metaheuristic algorithms for solving complex
optimization problems has necessitated robust and standardized
benchmarking practices. A fair and comprehensive evaluation is
crucial for validating an algorithm's performance, understanding
its strengths and weaknesses, and guiding future algorithmic
development. This paper provides a comprehensive review of the
landscape of benchmarking in metaheuristic optimization. We
systematically categorize and detail the primary types of
benchmark problems, including the classic set of 23 mathematical
functions, real-world engineering problems, specialized CEC
benchmark suites, combinatorial optimization benchmarks, and
multi-objective problems. For each category, we present the
fundamental mathematical formulations and discuss their specific
characteristics and challenges. Furthermore, we outline the
standard evaluation metrics used to quantify algorithmic
performance. Finally, we discuss current challenges in
benchmarking, such as the no free lunch theorem, the issue of
overfitting to test suites, and the need for more diverse and realworld
benchmarks. This review serves as a foundational guide for
researchers and practitioners in selecting appropriate
benchmarks and conducting rigorous, reproducible evaluations of
metaheuristic algorithms.

The examination of the relationship between energy use, reliability and network characteristics in water distribution systems (WDS) requires a set of sufficient case studies to give statistical significance to the conclusions. Considering the difficulty of finding real-world systems that can be used in this kind of research, synthetic (virtual) distribution systems are the next available alternative to use as case studies. This paper describes the use of a new method for the generation of synthetic distribution systems and its subsequent comparison against real-world systems to validate the suitability of the synthetic set to drive the conclusions of future research. The algorithm for the generation of synthetic WDSs is based on the work, methods and software presented by Mair et al. (2014). For the validation procedure, five metrics or indexes that account for the network's connectivity (link density, average node degree, meshedness coefficient) and the system's reliability (Resilience Index, Network Resilience) are evaluated; the ranges in each of the metrics are then compared. The generation of synthetic WDS required an enhancement of the connectivity of the networks and a pipe sizing that accounted for practical system design. An acceptable degree of similarity between the synthetic and the real-life sets of WDSs was achieved, although some modifications to the networks may be required in the future.

Remarkable advancements have been achieved in machine learning and computer vision through the utilization of deep neural networks. Among the most advantageous of these networks is the convolutional neural network (CNN). It has been used in pattern recognition, medical diagnosis, and signal processing, among other things. Actually, for these networks, the challenge of choosing hyperparameters is of utmost importance. The reason behind this is that as the number of layers rises, the search space grows exponentially. In addition, all known classical and evolutionary pruning algorithms require a trained or built architecture as input. During the design phase, none of them consider the process of pruning. In order to assess the effectiveness and efficiency of any architecture created, pruning of channels must be carried out before transmitting the dataset and computing classification errors. For instance, following pruning, an architecture of medium quality in terms of classification may transform into an architecture that is both highly light and accurate, and vice versa. There exist countless potential scenarios that could occur, which prompted us to develop a bi-level optimization approach for the entire process. The upper level involves generating the architecture while the lower level optimizes channel pruning. Evolutionary algorithms (EAs) have proven effective in bi-level optimization, leading us to adopt the co-evolutionary migration-based algorithm as a search engine for our bi-level architectural optimization problem in this research. Our proposed method, CNN-D-P (bi-level CNN design and pruning), was tested on the widely used image classification benchmark datasets, CIFAR-10, CIFAR-100 and ImageNet. Our suggested technique is validated by means of a set of comparison tests with regard to relevant state-of-the-art architectures.

Remarkable advancements have been achieved in machine learning and computer vision through the utilization of deep neural networks. Among the most advantageous of these networks is the convolutional neural network (CNN). It has been used in pattern recognition, medical diagnosis, and signal processing, among other things. Actually, for these networks, the challenge of choosing hyperparameters is of utmost importance. The reason behind this is that as the number of layers rises, the search space grows exponentially. In addition, all known classical and evolutionary pruning algorithms require a trained or built architecture as input. During the design phase, none of them consider the process of pruning. In order to assess the effectiveness and efficiency of any architecture created, pruning of channels must be carried out before transmitting the dataset and computing classification errors. For instance, following pruning, an architecture of medium quality in terms of classification may transform into an architecture that is both highly light and accurate, and vice versa. There exist countless potential scenarios that could occur, which prompted us to develop a bi-level optimization approach for the entire process. The upper level involves generating the architecture while the lower level optimizes channel pruning. Evolutionary algorithms (EAs) have proven effective in bi-level optimization, leading us to adopt the co-evolutionary migration-based algorithm as a search engine for our bi-level architectural optimization problem in this research. Our proposed method, CNN-D-P (bi-level CNN design and pruning), was tested on the widely used image classification benchmark datasets, CIFAR-10, CIFAR-100 and ImageNet. Our suggested technique is validated by means of a set of comparison tests with regard to relevant state-of-the-art architectures.

We explore the hypothesis that adding conservation of matter to an artificial life system can increase its evolutionary activity, through experiments with the Stringmol artificial chemistry. Our first experiment examines the effect of varying the number of opcodes and finds a concentration which maximises the evolutionary activity of the system. The second experiment searches for the optimum relative concentrations of opcodes that maximises evolutionary activity: it finds increased evolutionary activity, a high diversity of opcode concentrations in each search run, and a different configuration of concentrations in separate search runs. The third experiment investigates the need for low concentrations of opcodes in high evolutionary activity, and finds that evo activity decreass when more of these particular opcodes are provided. We conclude that conservation of matter provides an important evolutionary pressure that can lead to more diversity and more evolutionary activity, and is ...

In this paper we present a unified view of AI inspired by ideas from Evolutionary Computation as design of bounded rational agents. The approach specifies optimal programs rather than optimal actions, and is based on process algebras and anytime algorithms. The search method described in this paper is so general than many other search algorithms, including evolutionary search methods, become its special case. In this paper, we present a practical design of the programming language and environment targetting real-time complex domains. As AI systems move into more complex domains, all problems become real-time, because the agent will never have long enough time to solve the decision problem exactly.

It is interesting that typically in the proof of convergence of evolutionary algorithms only elitist selection is considered. In this paper, we stress out that truly in reaching optimum of the fitness function completeness of search plays probably even more important role. The elitist selection helps in reaching the optimum. It allows to converge faster and not to lose the optimum in cases when we are uncertain that the optimum has been reached. An evolutionary search using elitist selection, but being incomplete, will not reach the optimum. This paper provides sufficient conditions for finding the best (optimal) solutions, or the best (totally optimal) solutions with minimal search costs. To do that we utilitze a new formal model of evolutionary computation, the Evolutionary Turing Machine. The results are applicable both to genetic algorithms, genetic programming, evolution strategies and evolutionary programming. The problem of finding total optimum, optimizing together the quality of solution together with an evolutionary algorithm is considered as a multiobjective optimization allowing to tackle real-world problems where the complexity of evolutionary search becomes an issue. A new classification of evolutionary procedures into easy, hard, and solvable in the limit classes, is proposed.

We outline a theory of evolutionary computation using a formal model of evolutionary computation -the Evolutionary Turing Machine -which is introduced as the extension of the Turing Machine model. Evolutionary Turing Machines provide a better and a more complete model for evolutionary computing than conventional Turing Machines, algorithms, and Markov chains. The convergence and convergence rate are defined and investigated in terms of this new model. The sufficient conditions needed for the completeness and optimality of evolutionary search are investigated. In particular, the notion of the total optimality as an instance of the multiobjective optimization of the Universal Evolutionary Turing Machine is introduced. This provides an automatic way to deal with the intractability of evolutionary search by optimizing the quality of solutions and search costs simultaneously. Based on a new model a very flexible classification of optimization problem hardness for the evolutionary techniques is proposed. The expressiveness of evolutionary computation is investigated. We show that the problem of the best evolutionary algorithm is undecidable independently of whether the fitness function is time dependent or fixed. It is demonstrated that the evolutionary computation paradigm is more expressive than Turing Machines, and thus the conventional computer science based on them. We show that an Evolutionary Turing Machine is able to solve nonalgorithmically the halting problem of the Universal Turing Machine and, asymptotically, the best evolutionary algorithm problem. In other words, the best evolutionary algorithm does not exist, but it can be potentially indefinitely approximated using evolutionary techniques.

This research aims to improve real-time decision-making, process-state reconstruction, and multi-objective operational optimization in intelligent composting systems through an integrated framework based on hierarchical digital twin and Edge-Cloud architecture. Unlike previous studies that mainly focused on data monitoring or static optimization, the proposed framework enables dynamic reconstruction of process state, predictive decision-making, and intelligent assignment of tasks between Edge and Cloud layers simultaneously. The main innovation of the research lies in the combination of multilayer digital twin, dynamic decision rule for Edge-Cloud orchestration, and fuzzy multi-objective optimization in an integrated structure. In the proposed model, biological and operational uncertainties are modeled using triangular fuzzy numbers and control decisions are updated in real-time based on the actual system state. The results, compared to the baseline system without hierarchical digital twin and without Edge-Cloud orchestration, showed that the proposed framework was able to reduce the composting process time by about 28%, significantly reduce energy consumption, increase the compost quality index by 0.91, and effectively control the emission of undesirable compounds. The results also showed that the hierarchical Edge-Cloud architecture, by transferring time-sensitive decisions to the Edge layer and performing complex analyses in the Cloud, simultaneously improved the response time, process state reconstruction accuracy, and decision-making stability under dynamic and uncertain conditions. This research is an effective step in the development of intelligent, predictive, and self-adaptive systems for biological processes and sustainable waste management.

Management and analyses of water resources is of paramount importance in the implementation of water related sustainable development goals. Hydraulic models are key in flood forecasting and simulation applied to a river flood analysis and risk prediction and an accurate estimation of the roughness is one of the main factors in predicting the discharge in a stream. In practical implementation roughness can be represented by the prediction of the well known Manning's coefficient necessary for discharge calculation. In this paper we design an objective function that measures the quality of a given configuration of the Manning's coefficient. Such an objective function is optimised through several evolutionary approaches, namely: (1+1)-ES, CMA-ES, Differential Evolution, Particle Swarm Optimization and Bayesian Optimization. As case of study, a river in the central Italy was considered. The results indicate that the model, consistent with the classical techniques adopted in the hydraulic engineering field, is applicable to natural rivers and is able to provide an estimation of the roughness coefficients with a satisfactory accuracy. A comparison of the performances of the five evolutionary algorithms is also proposed.

Differential Evolution (DE) is a popular and efficient optimization technique for real-valued spaces based on the concepts of Darwinian evolution. Its main peculiarity is the use of a differential mutation operator that allows DE to automatically adjust the exploration/exploitation balance of its search moves. The major DE drawback is the need of a preliminary tuning of some numerical parameters. Although, recently some parameters adaptive schemes have been proposed, none of these takes into account the side effects introduced by changing two or more parameters at the same time. In this paper we introduce a DE self-adaptive scheme that takes into account the parameters dependencies by means of a multivariate probabilistic technique based on an Estimation of Distribution Algorithm working on the parameters space. Experiments have been performed on a set of commonly adopted benchmark problems and the obtained results show the competitiveness of our approach with respect to other adaptive DE schemes. Moreover, our scheme could be potentially adopted not only in DE but also in any other Evolutionary Algorithm or meta-heuristic technique presenting parameters that regulate the behavior of the search.

The GBP/USD currency pair, commonly known as "Cable," remains one of the most traded instruments in global foreign exchange markets. Every day, billions of dollars flow through this currency pair as banks, hedge funds, corporations, proprietary trading firms, and retail traders participate in price discovery. While the pair remains attractive because of its liquidity and volatility, the environment in which it trades has changed dramatically over the past decade. Today's markets are faster, more data-driven, and increasingly dominated by automation. Economic releases are analyzed within milliseconds, algorithmic systems react instantly to price changes, and liquidity providers constantly update quotes in response to changing market conditions. As a result, many traders who rely solely on manual execution are finding it increasingly difficult to compete. In fast-moving conditions, success depends not only on predicting market direction but also on execution speed, consistency, and disciplined risk management. Manual traders often struggle with these requirements because human decision-making introduces delays and emotional biases. Automated trading systems, by contrast, can process information and execute trades at speeds impossible for humans to match. This article examines why manual GBP/USD trading is becoming less effective in modern markets, compares manual and automated approaches, and explains how advanced risk management frameworks in automated trading help improve long-term performance. The Changing Nature of GBP/USD Markets

Evolutionary algorithms (EAs) can be used as optimisation mechanisms. Based on the model of natural evolution, they work on populations of individuals instead of on single solutions. In this way, the search is performed in a parallel manner. During the last decades, there has been a n increasing interest in evolutionary algorithms to solve scheduling problems. One important feature in these algorithms is the selection of individuals. Selection is the operation by which individuals (i.e. their chromosomes) are selected for mating. To emulate natural selection, individuals with higher fitness should be selected with higher probability, and thus it is one of the operators where the fitness plays an important role. There are many different models of selection (some are not biologically plausible). Commonly, proportional, ranking, tournament selection and stochastic universal sampling are used. EAs considered in this work are improved with a multiplicity feature to solve the job shop scheduling problems (JSSP). The algorithm applied here, multiple crossovers on multiple parents (MCMP), considers more than two parents for reproduction with the possibility to generate multiple children. This approach uses a permutation representation for the chromosome. The objective of this work is to compare the algorithms performance using different selection mechanisms and to analyse the different crossover methods developed to apply MCMP with a permutation representation.

Over the past few years, a continually increasing number of research efforts have investigated the application of evolutionary computation techniques for the solution of scheduling problems. Scheduling can pose extremely complex combinatorial optimization problems, which belong to the NP-hard family. Last enhancements on evolutionary algorithms include new multirecombinative approaches. Multiple Crossovers Per Couple (MCPC) allows multiple crossovers on the couple selected for mating and Multiple Crossovers on Multiple Parents (MCMP) do this but on a set of more than two parents. Techniques for preventing incest also help to avoid premature convergence. Issues on representation and operators in¯uence ef®ciency and ef®cacy of the algorithm. The present paper shows how enhanced evolutionary approaches, can solve the Job Shop Scheduling Problem (JSSP) in single and multiobjective optimization.

Evolutionary algorithms apply the processes of variation, reproduction and selection to look for an individual that is capable of solving the task in hand. In order to improve the evolvability of a population, we propose to copy important characteristics of nature's search space. Desired characteristics for a genotype-phenotype mapping are described, and several highly redundant genotype-phenotype mappings are analyzed in

This paper introduces an Automatic Wire Routing (AWR) algorithm based on evolutionary principles. In the proposed algorithm (dubbed EWRA, where "E" stands for evolutionary), populations of wiring networks evolve through the action of stochastic operators conceived from the heuristics of manual wire-routing. In each wiring configuration (individual of the population), all nets are routed and temporary intersections are allowed. The objective of evolution is to improve wiring quality in the sense of eliminating unwanted net intersections and simplifying paths in order to reduce manufacturing costs and increase reliability. Although the populations approach and the use of stochastic operators for simulating evolution resemble conventional genetic algorithms (GAS) operation, the AWR problem requires much more advanced representation and sophisticated operators whose action takes place at the wiring level, and not at the string level. Results showing the effectiveness of the proposed algorithm in a serial implementation are presented. Furthermore, in order to reduce computation time, parallel implementation of the proposed AWR algorithm is discussed for an ideal multicomputer architecture and for a realistic binary-tree multicomputer, in which the algorithm is currently being implemented. Preliminary results indicate that the proposed EWRA is feasible for AWR applications and suggest that efficient implementations can be obtained by using massively parallel processing. minals distnbuted through a wiring space. Although in this paper all the discussion will be for the single-layer case (two-dimensional wiring space), extension to the multilayer case is immediate. As the number of terminals to be connected and the wiring space increase, the search space increases in a non-polynomial fashion. In a common representation of the search space, the wiring space is viewed as a rectangular grid with mesh size A = w + c, where w is the minimum width of a net and c is the minimum clearance between nets. This representation is illustrated in Fig. , where two possible configurations are presented for nets connecting the terminals A-A', B-B', and C-C' on a single-layer grid. In Fig. l(a), the terminals C and C' cannot be connected. A possible wiring solution is presented in Fig. l(b), where it becomes clear that the order in which connections are made affects crucially the final results.

The paper presents an optimal fuzzy logic control algorithm for vibration mitigation of buildings using magneto-rheological (MR) dampers. MR dampers are semi-active devices and are monitored using external voltage supply. The voltage monitoring of MR damper is accomplished using evolutionary fuzzy system, where the fuzzy system is optimized using evolutionary algorithms (EAs). A micro-genetic algorithm (m-GA) and a particle swarm optimization (PSO) are used to optimize the FLC parameters. Two cases of optimal FLCs are shown. One where FLC is optimized keeping the rule base predefined and in the other case, FLC rule base is also optimized along with other FLC parameters. The FLC rule base and membership function parameters are optimized using 10 variables. Fuzzy controllers with a predefined rule base and with an optimal rule base are applied to a single degree of freedom (SDOF) and a multidegree of freedom (MDOF) system. Finally, the study evaluates the performance of the fuzzy controller optimized off-line, on a three storey building model under seismic excitations. The main advantage of using FLC to drive the MR damper voltage is that it provides a gradual and smooth change in voltage. Consequently, the present approach provides a better vibration control for structures under earthquake excitations.

Textile simulation models are notorious for being difficult to tune. The physically based derivations of energy functions, as mostly used for mapping the characteristics of real-world textiles on to simulation models, are labour-intensive and not guarantee satisfactory results. The extremely complex behaviour of textiles requires additional adjustment over a wide-range of parameters in order to achieve realistic reallife behaviour of the model. Furthermore, such derivations might not even be possible when dealing with mass-spring particle system-based models. Since there is no explicit correlation between the physical characteristics of textiles and the stiffnesses of springs that control a model's behaviour, this remains an unresolved issue. This paper proposes a hybrid evolutionary algorithm (EA), in order to solve this problem. The initial parameters of the model are written in individual's genes, where the number of genes is predefined for different textile types in order to limit the search-space. By mimicking the evolution processes, the EA is used to search the stability domain of the model to find a set of parameters that persuasively imitate the behaviour of a given real-world textile (e.g. silk, cotton or wool). This evaluation is based on the drape measurement, a characteristic often used when evaluating fabrics within the textile industry. The proposed EA is multi-objective, as textile drape is analysed using different quantifications. Local search is used to heuristically improve convergence towards a solution, while the efficiency of the method is demonstrated in comparison to a simple EA. To the best of our knowledge, this problem is being solved using an EA for the first time.

An Evolutionary Algorithm (EA) is a meta-heuristic and stochastic optimization search process that mimics Darwinian evolution theory and Mendel's Genetics. Each process facilitates (a) population(s) evolve into fittest and/or convergence by setting parameters of selection, mutation, crossover, population resizing, and/or many other variant operators. However, due to two primary identified factors, EAs are still a challenging research topic: (1) Value choices/ranges for parameters (i.e., parameter settings) will greatly influence the evolution performance of a search process in terms of fittest and/or convergence; and (2) Parameter settings that are good for one fitness function do not guarantee the same evolution performance of another fitness function. Namely, parameter settings are function-specific. Different functions may have various characteristics that request specific attention. In order to better organize and overcome the parameter setting problem, Eiben et al. have classified parameter settings into parameter tuning and parameter control : Parameter tuning determines parameter values before a search process begins while parameter control changes parameter values during a search process. More specifically, parameter control adjusts parameters on-the-fly using three different approaches: (1) Deterministic approach alters parameters based on certain pre-determined rules or formulae; (2) Adaptive approach strategically adjusts parameter values based on the feedbacks of a search process. Such feedbacks could be fitness, diversity, distance, among others; and (3) Self-adaptive approach encodes parameters to be adapted and evolves them along with a search process. Yet, even with such a classification, to our best knowledge there is no existing tool to assist researchers with conducting experiments of parameter settings with ease. Namely, researchers need to find out appropriate places out of thousand lines of EA source code to introduce and update specific parameters (including feedbacks) as well as formulae and adaptive strategies. Additionally, a number of revisions for EA source code will be also required for different kinds of experiments. To EA experimenters, such endeavor is time consuming and error prone. To EA developers, complex and tangling source code, resulted from different www.

Deep Neural Networks (DNN's) are a widely-used solution for a variety of machine learning problems. However, it is often necessary to invest a significant amount of a data scientist's time to pre-process input data, test different neural network architectures, and tune hyper-parameters for optimal performance. Automated machine learning (autoML) methods automatically search the architecture and hyper-parameter space for optimal neural networks. However, current state-of-the-art (SOTA) methods do not include traditional methods for manipulating input data as part of the algorithmic search space. We adapt the Evolutionary Multi-objective Algorithm Design Engine (EMADE), a multi-objective evolutionary search framework for traditional machine learning methods, to perform neural architecture search. We also integrate EMADE's signal processing and image processing primitives. These primitives allow EMADE to manipulate input data before ingestion into the simultaneously evolved DNN. We show that including these methods as part of the search space shows potential to provide benefits to performance on the CIFAR-10 image classification benchmark dataset.

Peningkatan nilai perkiraan akurasi dan interpretabilitas pada sebuah sistem samar adalah persoalan penting baik pada teori sistem samar ataupun pada aplikasinya. Telah diketahui bahwa optimisasi secara silmultan pada kedua persoalan tersebut adalah saling seimbang, namun hal ini akan meningkatkan pencapaian system dan menghindari pelatihan berlebihan. Particle Swarm Optimisasi (PSO) adalah bagian dari algoritma evolusioner yang merupakan calon algoritma yang baik untuk memecahkan masalah tersebut dan memiliki ruang pencarian global yang lebih baik. Tulisan ini mengenalkan sebuah integrasi antara PSO dan ANFIS untuk optimisasi pembelajarannya terutama pada penyesuaian nilai parameter fungsi kepemilikan dan menentukan jumlah aturan yang optimal untuk memperoleh nilai klasifikasi yang lebih baik. Usulan algoritma ini telah dites pada 4(empat) dataset standar dari mesin pembelajaran UCI, yaitu: dataset Iris Flower, Haberman's Survival Data, Balloon dan Thyroid. Hasil menunjukkan bahwa nilai klasifikasi yang lebih baik pada algoritma PSO-ANFIS yang diusulkan dan kompleksitas waktunya menurun bersesuaian.

In many pattern recognition tasks, an approach based on combining classifiers has shown a significant potential gain in comparison to the performance of an individual best classifier. This improvement turned out to be subject to a sufficient level of diversity exhibited among classifiers, which in general can be assumed as a selective property of classifier subsets. Given a large number of classifiers, an intelligent classifier selection process becomes a crucial issue of multiple classifier system design. In this paper, we have investigated three evolutionary optimization methods for the classifier selection task. Based on our previous studies of various diversity measures and their correlation with majority voting error we have adopted majority voting performance computed for the validation set directly as a fitness function guiding the search. To prevent from training data overfitting we extracted a population of best unique classifier combinations, and used them for second stage majority voting. In this work we intend to show empirically, that using efficient evolutionary-based selection leads to the results comparable to absolutely best, found exhaustively. Moreover, as we showed for selected datasets, introducing a second stage combining by majority voting has the potential for both, further improvement of the recognition rate and increase of the reliability of combined outputs.

A balanced diet is essential for a healthy life, and the food pyramid provides a simple visual guide to making good choices. The food pyramid is a graphic representation of nutritional recommendations, designed to guide people towards a healthy and balanced diet. It illustrates the types of foods and the amounts recommended for daily consumption, in order to maintain health and prevent nutrition-related diseases. The food pyramid is a visual guide on which a balanced diet is based. This representation is designed to determine the population to follow the dietary advice provided by a qualified medical organization. To interpret this schematization of the daily food ration, it has been agreed that the foods placed at the top are those that should be consumed in smaller quantities and with low frequency, and those placed at the bottom are those that should be consumed with greater frequency and in larger quantities. One of the fundamental principles of the food pyramid is proportion. It is essential to understand that the pyramid not only recommends certain types of foods, but also the proportion in which they should be consumed.

Metabolic processes in biological cells are commonly either characterized at the level of individual enzymes and metabolites or at the network level. Often these two paradigms are considered as mutually exclusive because concepts from neither side are suited to describe the complete range of scales. Additionally, when modeling metabolic or regulatory cellular systems, often a large fraction of the required kinetic parameters are unknown. This even applies to such simple and extensively studied systems like the photosynthetic apparatus of purple bacteria. Using the chromatophore vesicles of Rhodobacter sphaeroides as a model system, we show that a consistent kinetic model emerges when fitting the dynamics of a molecular stochastic simulation to a set of time dependent experiments even though about two thirds of the kinetic parameters in this system are not known from experiment. Those kinetic parameters that were previously known all came out in the expected range. The simulation model was built from independent protein units composed of elementary reactions processing single metabolites. This pools-and-proteins approach naturally compiles the wealth of available molecular biological data into a systemic model and can easily be extended to describe other systems by adding new protein or nucleic acid types. The automated parameter optimization, performed with an evolutionary algorithm, reveals the sensitivity of the model to the value of each parameter and the relative importances of the experiments used. Such an analysis identifies the crucial system parameters and guides the setup of new experiments that would add most knowledge for a systemic understanding of cellular compartments. The successful combination of the molecular model and the systemic parametrization presented here on the example of the simple machinery for bacterial photosynthesis shows that it is actually possible to combine molecular and systemic modeling. This framework can now straightforwardly be applied to other currently less well characterized but biologically more relevant systems.

The automated recognition of emotions from speech is a challenging issue. In order to build an emotion recognizer well defined features and optimized parameter sets are essential. This paper will show how an optimal parameter set for HMM-based recognizers can be found by applying an evolutionary algorithm on standard features in automated speech recognition. For this, we compared different signal features, as well as several architectures of HMMs. The system was evaluated on a non-acted database and its performance was compared to a baseline system. We present an optimal feature set for the public part of the SmartKom database.

Human beings are greatly inspired by nature. Nature has the ability to solve very complex problems in its own distinctive way. The problems around us are becoming more and more complex in the real time and at the same instance our mother nature is guiding us to solve these natural problems. Nature gives some of the logical and effective ways to find solutions to these problems. Nature acts as an optimized source for solving the complex problems. Decomposition is a basic strategy in traditional multi-objective optimization. However, it has not yet been widely used in multi-objective evolutionary optimization. Although computational strategies for taking care of Multi-objective Optimization Problems (MOPs) have been accessible for a long time, the ongoing utilization of Evolutionary Algorithm (EAs) to such issues gives a vehicle to tackle extremely enormous scope MOPs. MOBATD is a multi-objective bat algorithm that incorporates the dominance concept with the decomposition approach. Whilst decomposition simplifies the MOP by rewriting it as a set of Tchebycheff Approach, solving these problems simultaneously, within the BAT framework, might lead to premature convergence because of the leader selection process which uses the Tchebycheff Approach as a criterion. Dominance plays a major role in building the leaders archive, allowing the selected leaders to cover less dense regions while avoiding local optima and resulting in a more diverse approximated Pareto front. The results from 5 standard MOPs show that the MOBATD outperforms some developmental methods based on decomposition. All the results were achieved by MATLAB (R2017b).

This research shows the usefulness of fuzzy logic approaches for modelling and simulation of complex dynamical systems. Several hybrid soft computing methodologies based on fuzzy logic, such are neurofuzzy systems, genetic-fuzzy systems and the Fuzzy Inductive Reasoning are applied to a real dynamical system in the ecological domain, i.e. the global temperature change. The ocean-atmosphere system is represented in this work by using an energy balance model that reproduces a range of temperatures increase that agrees with that reported by the IPCC. The results obtained by all the fuzzy approaches studied are good, although the Fuzzy Inductive Reasoning methodology performs clearly much better that the other approaches for the application studied from the prediction accuracy point of view.

The losses in electrical power systems are a significant problem. The proper adjusting of reactive power resources is one of the ways for minimizing the Power Loss (P_L) in any power system. Reactive Power Optimization (RPO) is recorded as a complex optimization problem. The calculations of this problem are a part of Optimal Power Flow (OPF) calculations. In this work , a new algorithm called Weight Improved Particle Swarm Optimization (WIPSO) algorithm is presented on enhancing the function of weight parameters for solving this problem. Weight Improved Particle Swarm Optimization is presented as a useful optimization tool to search for optimal settings of reactive power independent variables during dealing with a number of equality, and inequality constraints at same time by minimizing the goal function (P_L). The Weight Improved Particle Swarm Optimization is tested on the IEEE 14-bus system. Simulation results obtained WIPSO was effective, and attain the best results and has best convergence characteristic and performance in terms of decreasing (P_L) compared to simple (PSO) algorithm .

Rocket-based combined cycle (RBCC) engines are an airbreathing propulsion technology that offers considerable potential for efficient access-to-space. Successful design of RBCC-powered space transport systems requires reliable databases for both vehicle and engine performance, calling for an effective sampling method to accurately resolve non-linear characteristics in vast design space. This paper presents an optimal sampling strategy based on the function gradients to realize efficient database construction based on evolutionary algorithms and assesses its effectiveness by applying the methodology to various test functions with multiple objectives as well as surrogate models representing scramjet intake characteristics for validation.

This paper presents a minimal validated framework for designing energy-aware metaheuristics that operate under fixed energy budgets. We introduce a unified operator-level model that quantifies both numerical gain and energy consumption, and define a robust Expected Improvement per Joule (EI/J) score to guide adaptive selection among operator variants during the search. The resulting energy-aware solvers dynamically choose between operators to selfcontrol exploration and exploitation, aiming to maximise fitness gain under limited energy. We instantiate this framework in three representative metaheuristics-steady-state GA, PSO, and ILS-each equipped with two lightweight/heavy update variants in a controlled setting. Experiments on three heterogeneous combinatorial problems (Knapsack, NK-landscapes, and Error-Correcting Codes) show that the energy-aware variants can reach comparable fitness while requiring substantially less energy than their non-energyaware baselines. EI/J values stabilise early and yield clear operatorselection patterns, with each solver reliably self-identifying the most improvement-per-Joule-efficient operator across problems. CCS Concepts • Computing methodologies → Bio-inspired approaches; • Theory of computation → Evolutionary algorithms; • Social and professional topics → Sustainability.

Solving complex optimization problems increasingly relies on surrogate models to approximate expensive objective functions. While surrogate accuracy is often emphasized, other practical factors such as computational cost and resource consumption are usually overlooked. This paper introduces the PTME methodology, a structured approach to evaluate surrogate models by jointly considering Precision, Time, Memory, and Energy. We argue that combining numerical accuracy with physical performance metrics is essential to assess whether a surrogate can serve as both a reliable scientific proxy and an efficient computational tool. PTME is illustrated on a real-world traffic-light optimization problem in European cities, where deep learning surrogates are analyzed with respect to sampling strategy, dataset size, and resource usage during training and inference. Finally, an illustrative surrogate-assisted optimization experiment shows how PTME-based surrogate characterization translates into optimization behavior, highlighting trade-offs between solution quality and computational efficiency. Overall, PTME provides a practical guideline for transparent surrogate evaluation in resource-aware optimization studies.

In this paper we present a distributed Evolutionary Algorithm (EA) whose population is structured using newscast, a gossiping protocol. This algorithm has been designed to deal with computationally expensive problems via massive scalability; therefore, we analyse the response time of the model using large instances of well-known hard optimization problems that require from EAs a (sometimes exponentially) bigger computational effort as these problems scale. Our approach has been matched against a sequential Genetic Algorithm (sGA) applied to the same set of problems, and we found that it needs less computational effort than the sGA in yielding success. Furthermore, the response time scale logarithmically with respect to the problem size, which makes it suitable to tackle large instances.

Solving complex optimization problems increasingly relies on surrogate models to approximate expensive objective functions. While surrogate accuracy is often emphasized, other practical factors such as computational cost and resource consumption are usually overlooked. This paper introduces the PTME methodology, a structured approach to evaluate surrogate models by jointly considering Precision, Time, Memory, and Energy. We argue that combining numerical accuracy with physical performance metrics is essential to assess whether a surrogate can serve as both a reliable scientific proxy and an efficient computational tool. PTME is illustrated on a real-world traffic-light optimization problem in European cities, where deep learning surrogates are analyzed with respect to sampling strategy, dataset size, and resource usage during training and inference. Finally, an illustrative surrogate-assisted optimization experiment shows how PTME-based surrogate characterization translates into optimization behavior, highlighting trade-offs between solution quality and computational efficiency. Overall, PTME provides a practical guideline for transparent surrogate evaluation in resource-aware optimization studies.

This paper presents a minimal validated framework for designing energy-aware metaheuristics that operate under fixed energy budgets. We introduce a unified operator-level model that quantifies both numerical gain and energy consumption, and define a robust Expected Improvement per Joule (EI/J) score to guide adaptive selection among operator variants during the search. The resulting energy-aware solvers dynamically choose between operators to selfcontrol exploration and exploitation, aiming to maximise fitness gain under limited energy. We instantiate this framework in three representative metaheuristics-steady-state GA, PSO, and ILS-each equipped with two lightweight/heavy update variants in a controlled setting. Experiments on three heterogeneous combinatorial problems (Knapsack, NK-landscapes, and Error-Correcting Codes) show that the energy-aware variants can reach comparable fitness while requiring substantially less energy than their non-energyaware baselines. EI/J values stabilise early and yield clear operatorselection patterns, with each solver reliably self-identifying the most improvement-per-Joule-efficient operator across problems. CCS Concepts • Computing methodologies → Bio-inspired approaches; • Theory of computation → Evolutionary algorithms; • Social and professional topics → Sustainability.

At present, evolutionary optimization algorithms are increasingly used in the development of new technological processes. Evolutionary algorithms often allow the optimization procedure to be performed even in cases where classical optimization algorithms fail (e.g. gradient methods) and where an acceptable solution is sufficient to solve the optimization task. The article focuses on possibilities of using a differential evolution algorithm in the optimization process. This algorithm is often referred to in the literature as a global optimization procedure. However, we show by means of a practical example that the convergence of the classic differential algorithm to the global extreme is not generally assured and is largely dependent on the specific cost function. To remove this weakness, we designed a modified version of the differential evolution algorithm. The improved version, named the modified differential evolution algorithm, is described in the article. It is possible to prov...

Given a point in $m$-dimensional objective space, any $\varepsilon$-ball of a point can be partitioned into the incomparable, the dominated and dominating region. The ratio between the size of the incomparable region, and the dominated (and dominating) region decreases proportionally to $1/2^{m-1}$, i.e., the volume of the Pareto dominating orthant as compared to all other volumes. Due to this reason, it gets increasingly unlikely that dominating points can be found by random, isotropic mutations. As a remedy to stagnation of search in many objective optimization, in this paper, we suggest to enhance the Pareto dominance order by involving an obtuse convex dominance cone in the convergence phase of an evolutionary optimization algorithm. We propose edge-rotated cones as generalizations of Pareto dominance cones for which the opening angle can be controlled by a single parameter only. The approach is integrated in several state-of-the-art multi-objective evolutionary algorithms (MOEA...

In this paper we propose a Diversity-Indicator based Multi-Objective Evolutionary Algorithm (DI-MOEA) for fast computation of evenly spread Pareto front approximations. Indicator-based optimization has been a successful principle for multi-objective evolutionary optimization algorithm (MOEA) design. The idea is to guide the search for approximating the Pareto front by a performance indicator. Ideally, the indicator captures both convergence to the Pareto front and a high diversity, and it does not require a priori knowledge of the Pareto front shape and location. It is, however, so far difficult to define indicators that scale well in computation time for high dimensional objective spaces, and that distribute points evenly on the Pareto front. Moreover, the behavior of commonly applied indicators depends on additional information, such as reference points or sets. The proposed DI-MOEA adopts a hybrid search scheme for combining the advantages of Pareto dominance-based approaches to ensure fast convergence to the Pareto front, with indicator based approaches to ensure convergence to an evenly distributed, diverse set. In addition, it avoids the use of complex structure and parameters in decomposition-based approaches. The Euclidean distance-based geometric mean gap is used as diversity indicator. Experimental results show that the new algorithm can find uniformly spaced Pareto fronts without the involvement of any reference points or sets. Most importantly, our algorithm performs well on both the hypervolume indicator and IGD when comparing with state-of-the-art MOEAs (NSGA-II, SMS-EMOA, MOEA/D and NSGA-III).

In almost no other field of computer science, the idea of using bio-inspired search paradigms has been so useful as in solving multiobjective optimization problems. The idea of using a population of search agents that collectively approximate the Pareto front resonates well with processes in natural evolution, immune systems, and swarm intelligence. Methods such as NSGA-II, SPEA2, SMS-EMOA, MOPSO, and MOEA/D became standard solvers when it comes to solving multiobjective optimization problems. This tutorial will review some of the most important fundamentals in multiobjective optimization and then introduce representative algorithms, illustrate their working principles, and discuss their application scope. In addition, the tutorial will discuss statistical performance assessment. Finally, it highlights recent important trends and closely related research fields. The tutorial is intended for readers, who want to acquire basic knowledge on the mathematical foundations of multiobjective optimization and state-of-the-art methods in evolutionary multiobjective optimization. The aim is to provide a starting point for researching in this active area, and it should also help the advanced reader to identify open research topics.

Ricardo's famous comparative-advantage trade theory was built upon a production possibility curve (PPC), which contains only 1 input. Pareto tried to surpass Ricardo by developing a 2-input PPC, called contract curve or Pareto optimality. However, a 2-input production function is a 3-D matter. Pareto's 2-D presentation is thus inaccurately misleading. This paper uses the cross-sectional image of the 3-D model to disprove Pareto's contract curve, Pareto optimality and hence PPC. By Pareto's way, Ricardo's PPC and trade theory are also disproved.

This dissertation introduces a methodology and genetic encoding to optimize NARS configurations using evolutionary algorithms. In contrast to neural models optimized via gradients, NARS configuration spans symbolic knowledge structures, discrete mechanism choices, and continuous parameters, yielding objectives that are not readily differentiable; this motivates the use of evolutionary search. The proposed evolutionary approach serves a role analogous to pre-training in neural systems: it produces a task-adapted initialization of background knowledge and system configuration prior to deployment to achieve better performance and learning abilities. Given an agent’s sensorimotor vocabulary (i.e., the sets of its sensors and motors), the encoding generates and mutates background knowledge (“instincts”) intended to bootstrap useful behavior. In addition, the encoding optimizes numeric meta-parameters that influence NARS architecture and learning dynamics. To increase expressive power, the encoding was extended with support for the implantation and learning of variables and compound structures, improving performance in scenarios that benefit from more expressive knowledge patterns.