Design Of Evolutionary Algorithms Research Articles

Fuzzy logic applied to mutation size in evolutionary strategies

AbstractTuning of algorithm parameters is a complex but very important issue in the design of Evolutionary Algorithms. This paper discusses a new concept of mutation size tuning in Evolutionary Strategies. The proposed algorithm uses data on evolutionary history in earlier generations to tune the mutation size. A Fuzzy Logic Part examines this historical data and tunes the mutation size of individuals to improve the algorithm’s convergence and its resistance to getting stuck in a local optimum. The Fuzzy Logic Part tunes the mutation size and keeps an appropriate relation of algorithm’s exploration and exploitation. The proposed concept is discussed, and several tests on Function Optimization Problems are performed. In tests, we use a set of data and functions with different difficulties recommended in the commonly used benchmarks. The results of experiments suggest that the proposed method is more efficient and resistant to getting stuck in suboptimal solutions. The proposed algorithm has been used in recognizing the type of ultra-high energy cosmic ray particle that initiates the Extensive Air Showers when hit the Earth atmosphere. It could be used for a wide range of similar problems. It is possible that the proposed method could be adapted to other types of optimization methods, inspired by natural evolution, for example, Evolutionary Algorithms.

Learning to Learn Evolutionary Algorithm: A Learnable Differential Evolution

Research on evolutionary optimization has flourished for several decades. Now it has come to a turning point. With the advancement of artificial intelligence, especially deep learning and reinforcement learning, it is becoming appealing to rethink the design and development of evolutionary algorithm (EA). From our perspective, a new-generation EA should be <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">learned</i> rather than manually designed, based on learning from optimization experiences (such as obtained from optimizing a family of optimization problems), the deep understanding of the roles of recombination operators, and the usage of experiences extracted through history optimization trajectories, so as to intelligently decide the control parameters that can adapt to the problem landscape changes. This learning can be conducted by strongly coupling with reinforcement learning since an evolutionary search procedure can be modeled as a Markov Decision Process (MDP). In this paper we propose a framework for automatic learning of EA, and present an exemplar study on learning a differential evolution (DE). Experimental results show that the <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">learned</i> adaptive DE is very competitive to some recent EAs on a commonly-used test suite, which indicates that the proposed learning framework has a great potential for the <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">automatic</i> design of promising EAs.

Relation Between Objective Space Normalization and Weight Vector Scaling in Decomposition-Based Multiobjective Evolutionary Algorithms

Real-world multiobjective optimization problems (MOPs) usually have conflicting and differently-scaled objectives. To deal with such problems, objective space normalization is widely used in multiobjective evolutionary algorithm (MOEA) design, especially, in the design of decomposition-based MOEAs. It has been demonstrated that uniformly-distributed solutions can be obtained for badly-scaled MOPs by decomposition-based MOEAs with objective space normalization. Recently, weight vector scaling has also been used for badly-scaled MOPs. In some studies, it was argued that weight vector scaling and objective space normalization are essentially the same when applied to decomposition-based MOEAs. In this paper, we theoretically and empirically show the relation between objective space normalization and weight vector scaling. Our results demonstrate that similarities and differences between the two methods depend on the choice of a scalarizing function. How the choice between normalization and weight vector scaling affects decomposition-based MOEAs with solution assignment mechanisms is also analyzed.

Bi-EvoGAN: Bi-level Evolutionary Approach for Generative Adversarial Networks

Generative Adversarial Networks (GANs) are an adversarial approach to generative modeling using deep learning methods, which have become one of the most relevant topics in the last years. Despite producing interesting outputs, recent studies have demonstrated that GANs remain difficult to control and suffer from training problems, such as instability and mode collapse, as well as several aspects of their models are designed by hand. Neuroevolution is a recent evolution strategy that uses Evolutionary Algorithms (EAs) in order to provide an automatic design of deep neural networks architectures. Many recent efforts have been dedicated to contributing on GANs by developing different single-level EAs approaches, however they could not solve their common issues completely which are still open problems. In this paper, we propose a new design of evolutionary algorithms in the context of GANs, consisting of dividing the evolution search space into two complementary levels, modeling a Bi-level Evolutionary Approach called Bi-EvoGAN searching to surpass their adversarial optimization difficulties and also to demonstrate the performance of considering two levels of optimization in comparison to a single optimization level. To the best of our knowledge, this work presents the first application of the bi-level optimization in the context of generative adversarial networks. The upper level uses an Evolutionary Algorithm-based Topology Optimization method named “EA-TO” to evolve a population of GAN networks as a set of individuals in order to find the best network topology. The lower level applies an Evolutionary Algorithm-based Decomposition for Hyperparameter Optimization termed “EA/D-HO” in order to select the best hyperparameters of each GAN individual. The performance of Bi-EvoGAN is explained by the double optimization of the GAN topology and their hyperparameters through the new bilevel optimization design allowing to evolve a GAN population in its higher fitness which will improve the quality of the search method and consequently to accelerate the convergence. Experimental studies are carried out based on four well-known benchmark datasets, MNIST, Fashion-MNIST, CelebA and LSUN-Bedroom allowing to show the efficiency of our approach in comparison to the other state-of-the-art approaches.

Learning Adaptive Differential Evolution by Natural Evolution Strategies

Adaptive parameter control is critical in the design and application of evolutionary algorithm (EA), so does in differential evolution. In the past decade, many adaptive evolutionary algorithms have been proposed, in which online information collected until current generation during the evolutionary search procedure is used to determine the algorithmic parameters for the next generation. Recent studies often assume that the algorithmic parameters follow some distributions, while the distributions' parameters (called hyper-parameters) are updated by the collected information. Performances of these adaptive EAs depend highly on the hyper-parameters. Notice that the experiences obtained from optimizing some related problems could provide useful guidelines on how to adaptively control the distributions' parameters. However, few existing studies sufficiently used such experiences. To fill the gap, we propose a general framework for adaptive parameter control by modeling its evolution procedure as a Markov decision process. In the framework, a neural network is employed to act as the controller. The natural evolution strategies is applied to train the neural network. The proposed framework is applied on two well-known differential evolutions (DEs), namely JADE and LSHADE. By incorporating the learned controller, two DEs, named JADE/AC and LSHADE/AC, are formed. Experimental results on the CEC 2018 benchmark suite show that in general JADE/AC and LSHADE/AC perform significantly better than their counterparts. Moreover, in comparison with some well-known EAs including three suggested best DEs in a review paper (including LSHADE, cDE and CoBiDE), the championship algorithm in the CEC 2018 competitions, a recently-developed learnable DE and recently proposed DEs, our study shows that LSHADE/AC performs the best amongst them without sacrificing much computation time.

Real-time data-driven automatic design of multi-objective evolutionary algorithm: A case study on production scheduling

Real-time data-driven automatic design of multi-objective evolutionary algorithm: A case study on production scheduling

An Evolutionary Algorithm with Gene Ontology-Aware Crossover Operator for Protein Complex Detection

Evolutionary algorithms (EAs), as global search methods, are proved to be more robust than their counterpart local heuristics for detecting protein complexes in protein-protein interaction (PPI) networks. Typically, the source of robustness of these EAs comes from their components and parameters. These components are solution representation, selection, crossover, and mutation. Unfortunately, almost all EA based complex detection methods suggested in the literature were designed with only canonical or traditional components. Further, topological structure of the protein network is the main information that is used in the design of almost all such components. The main contribution of this paper is to formulate a more robust EA with more biological consistency. For this purpose, a new crossover operator is suggested where biological information in terms of both gene semantic similarity and protein functional similarity is fed into its design. To reflect the heuristic roles of both semantic and functional similarities, this paper introduces two gene ontology (GO) aware crossover operators. These are direct annotation-aware and inherited annotation-aware crossover operators. The first strategy is handled with the direct gene ontology annotation of the proteins, while the second strategy is handled with the directed acyclic graph (DAG) of each gene ontology term in the gene product. To conduct our experiments, the proposed EAs with GO-aware crossover operators are compared against the state-of-the-art heuristic, canonical EAs with the traditional crossover operator, and GO-based EAs. Simulation results are evaluated in terms of recall, precision, and F measure at both complex level and protein level. The results prove that the new EA design encourages a more reliable treatment of exploration and exploitation and, thus, improves the detection ability for more accurate protein complex structures.

The Importance of Diversity in the Variable Space in the Design of Multi-Objective Evolutionary Algorithms

Most current Multi-Objective Evolutionary Algorithms (moeas) do not directly manage the population’s diversity in the variable space. Usually, these kind of mechanisms are only considered in Evolutionary Multimodal Multi-Objective Algorithms (emmas) which aim to obtain a complete representation of the set of – locally or globally – optimal solutions in variable space. This is a remarkable difference with respect to single-objective optimizers, where maintaining diverse solutions is considered favorable to better explore the search space. The contribution of this research is to show that the quality of current moeas in terms of objective space metrics can be enhanced by integrating mechanisms to explicitly manage the diversity in the variable space. The key is to consider the stopping criterion and elapsed period in order to dynamically alter the importance granted to the diversity in the variable space and to the quality and diversity in the objective space, which is an important difference with respect to emmas. Specifically, more importance is given to the variable space in the initial phases and, decisions are progressively more biased by the information of the objective space as the evolution progresses. This paper presents a novel moea based on decomposition (avsd-moea/d) that relies on these principles by means of a novel replacement phase. Extensive experimentation shows the clear benefits provided by the proposed design principle.

Design of Evolutionary Algorithm Based Energy Efficient Clustering Approach for Vehicular Adhoc Networks

In a vehicular ad hoc network (VANET), a massive quantity of data needs to be transmitted on a large scale in shorter time durations. At the same time, vehicles exhibit high velocity, leading to more vehicle disconnections. Both of these characteristics result in unreliable data communication in VANET. A vehicle clustering algorithm clusters the vehicles in groups employed in VANET to enhance network scalability and connection reliability. Clustering is considered one of the possible solutions for attaining effectual interaction in VANETs. But one such difficulty was reducing the cluster number under increasing transmitting nodes. This article introduces an Evolutionary Hide Objects Game Optimization based Distance Aware Clustering (EHOGO-DAC) Scheme for VANET. The major intention of the EHOGO-DAC technique is to portion the VANET into distinct sets of clusters by grouping vehicles. In addition, the DHOGO-EAC technique is mainly based on the HOGO algorithm, which is stimulated by old games, and the searching agent tries to identify hidden objects in a given space. The DHOGO-EAC technique derives a fitness function for the clustering process, including the total number of clusters and Euclidean distance. The experimental assessment of the DHOGO-EAC technique was carried out under distinct aspects. The comparison outcome stated the enhanced outcomes of the DHOGO-EAC technique compared to recent approaches.

Design of Evolutionary Algorithm Based Unequal Clustering for Energy Aware Wireless Sensor Networks

Wireless Sensor Networks (WSN) play a vital role in several real-time applications ranging from military to civilian. Despite the benefits of WSN, energy efficiency becomes a major part of the challenging issue in WSN, which necessitate proper load balancing amongst the clusters and serves a wider monitoring region. The clustering technique for WSN has several benefits: lower delay, higher energy efficiency, and collision avoidance. But clustering protocol has several challenges. In a large-scale network, cluster-based protocols mainly adapt multi-hop routing to save energy, leading to hot spot problems. A hot spot problem becomes a problem where a cluster node nearer to the base station (BS) tends to drain the energy much quicker than other nodes because of the need to implement more transmission. This article introduces a Jumping Spider Optimization Based Unequal Clustering Protocol for Mitigating Hotspot Problems (JSOUCP-MHP) in WSN. The JSO algorithm is stimulated by the characteristics of spiders naturally and mathematically modelled the hunting mechanism such as search, persecution, and jumping skills to attack prey. The presented JSOUCP-MHP technique mainly resolves the hot spot issue for maximizing the network lifespan. The JSOUCP-MHP technique elects a proper set of cluster heads (CHs) using average residual energy (RE) to attain this. In addition, the JSOUCP-MHP technique determines the cluster sizes based on two measures, i.e., RE and distance to BS (DBS), showing the novelty of the work. The proposed JSOUCP-MHP technique is examined under several experiments to ensure its supremacy. The comparison study shows the significance of the JSOUCP-MHP technique over other models.

Improving quantum genetic optimization through granular computing

Quantum computers promise to revolutionize the world of computing thanks to some features of quantum mechanics that can enable massive parallelism in computation. This benefit may be particularly relevant in the design of evolutionary algorithms, where the quantum paradigm could support the exploration of multiple regions of the search space in a concurrent way. Although some efforts in this research field are ongoing, the potential of quantum computing is not yet fully expressed due to the limited number of qubits of current quantum processors. This limitation is even more acute when one wants to deal with continuous optimization problems, where the search space is potentially infinite. The goal of this paper is to address this limitation by introducing a hybrid and granular approach to quantum algorithm design, specifically designed for genetic optimization. This approach is defined as hybrid, because it uses a digital computer to evaluate fitness functions, and a quantum processor to evolve the genetic population; moreover, it uses granular computing to hierarchically reduce the size of the search space of a problem, so that good near-optimal solutions can be identified even on small quantum computers. As shown in the experiments, where IBM Q family processors are used, the usage of a granular computation scheme statistically enhances the performance of the state-of-the-art evolutionary algorithm implemented on quantum computers, when it is run to optimize well-known benchmark continuous functions.

Cultural Algorithms as a Framework for the Design of Trustable Evolutionary Algorithms

One of the major challenges facing Artificial Intelligence in the future is the design of trustworthy algorithms. The development of trustworthy algorithms will be a key challenge in Artificial Intelligence for years to come. Cultural Algorithms (CAs) are viewed as one framework that can be employed to produce a trustable evolutionary algorithm. They contain features to support both sustainable and explainable computation that satisfy requirements for trustworthy algorithms proposed by Cox [Nine experts on the single biggest obstacle facing AI and algorithms in the next five years, Emerging Tech Brew, January 22, 2021]. Here, two different configurations of CAs are described and compared in terms of their ability to support sustainable solutions over the complete range of dynamic environments, from static to linear to nonlinear and finally chaotic. The Wisdom of the Crowds method was selected for the one configuration since it has been observed to work in both simple and complex environments and requires little long-term memory. The Common Value Auction (CVA) configuration was selected to represent those mechanisms that were more data centric and required more long-term memory content. Both approaches were found to provide sustainable performance across all the dynamic environments tested from static to chaotic. Based upon the information collected in the Belief Space, they produced this behavior in different ways. First, the topologies that they employed differed in terms of the “in degree” for different complexities. The CVA approach tended to favor reduced “indegree/outdegree”, while the WM exhibited a higher indegree/outdegree in the best topology for a given environment. These differences reflected the fact the CVA had more information available for the agents about the network in the Belief Space, whereas the agents in the WM had access to less available knowledge. It therefore needed to spread the knowledge that it currently had more widely throughout the population.

Single Red Blood Cell Hydrodynamic Traps via the Generative Design.

This paper describes a generative design methodology for a micro hydrodynamic single-RBC (red blood cell) trap for applications in microfluidics-based single-cell analysis. One key challenge in single-cell microfluidic traps is to achieve desired through-slit flowrates to trap cells under implicit constraints. In this work, the cell-trapping design with validation from experimental data has been developed by the generative design methodology with an evolutionary algorithm. L-shaped trapping slits have been generated iteratively for the optimal geometries to trap living-cells suspended in flow channels. Without using the generative design, the slits have low flow velocities incapable of trapping single cells. After a search with 30,000 solutions, the optimized geometry was found to increase the through-slit velocities by 49%. Fabricated and experimentally tested prototypes have achieved 4 out of 4 trapping efficiency of RBCs. This evolutionary algorithm and trapping design can be applied to cells of various sizes.

Decoding the Architectural Genome: Multi-Objective Evolutionary Algorithms in Design

The application of population-based optimization algorithms in design is heavily driven by the translation and analysis of various data sets that represent a design problem; in evolutionary-based algorithms, these data sets are illustrated through two primary data streams: genes and fitness functions. The latter is frequently examined when analyzing the algorithm’s output, and the former is comparatively less so. This paper examines the role of genomic analysis in applying multi-objective evolutionary algorithms (MOEA) in design. The results demonstrate the significance of utilizing the genetic analysis to understand better the relationships between parameters used in the design problem’s formulation and differentiate between morphological differences in the algorithmic output not commonly observed through fitness-based analyses.

Improved Lebesgue Indicator-Based Evolutionary Algorithm: Reducing Hypervolume Computations

One of the major limitations of evolutionary algorithms based on the Lebesgue measure for multi-objective optimization is the computational cost required to approximate the Pareto front of a problem. Nonetheless, the Pareto compliance property of the Lebesgue measure makes it one of the most investigated indicators in the design of indicator-based evolutionary algorithms (IBEAs). The main deficiency of IBEAs that use the Lebesgue measure is their computational cost which increases with the number of objectives of the problem. On this matter, the investigation presented in this paper introduces an evolutionary algorithm based on the Lebesgue measure to deal with box-constrained continuous multi-objective optimization problems. The proposed algorithm implicitly uses the regularity property of continuous multi-objective optimization problems that has suggested effectiveness when solving continuous problems with rough Pareto sets. On the other hand, the survival selection mechanism considers the local property of the Lebesgue measure, thus reducing the computational time in our algorithmic approach. The emerging indicator-based evolutionary algorithm is examined and compared versus three state-of-the-art multi-objective evolutionary algorithms based on the Lebesgue measure. In addition, we validate its performance on a set of artificial test problems with various characteristics, including multimodality, separability, and various Pareto front forms, incorporating concavity, convexity, and discontinuity. For a more exhaustive study, the proposed algorithm is evaluated in three real-world applications having four, five, and seven objective functions whose properties are unknown. We show the high competitiveness of our proposed approach, which, in many cases, improved the state-of-the-art indicator-based evolutionary algorithms on the multi-objective problems adopted in our investigation.

PEAB: A pool-based distributed evolutionary algorithm model with buffer

Pool Model is an asynchronous, loosely coupled distributed evolutionary algorithm (dEA) design architecture. However, the classical Pool Model face some design problems, such as population control, work redundancy, rough selection/replacement strategies, and unreliable connections, etc. In this paper, a novel distributed pool evolutionary algorithm (EA) model with buffer (PEAB) is proposed. PEAB can solve the inherent problems of the Pool Model by using the buffer setting, the Reunion mechanism, and the Migration in Pool (MP) strategy. Besides, PEAB provides stronger population control and more global population selection/replacement strategies. In the experimental part, we compared PEAB with another Pool Model named EvoSpace using a common benchmark. The experiments showed that the convergence rate of PEAB is 59.7% faster than that of EvoSpace under the respective fastest conditions. PEAB also has a faster reception rate of the first generation and stronger population control. Besides, this paper also tests and analyzes the scalability of PEAB using two other benchmarks. The overall trend of the experiment results suggested that PEAB would be faster with more Workers. Last but not least, this paper studies the effect of the MP strategy on the performance of PEAB, and the results showed that the MP strategy can effectively improve the convergence efficiency.

A Comparison of Archiving Strategies for Characterization of Nearly Optimal Solutions under Multi-Objective Optimization

In a multi-objective optimization problem, in addition to optimal solutions, multimodal and/or nearly optimal alternatives can also provide additional useful information for the decision maker. However, obtaining all nearly optimal solutions entails an excessive number of alternatives. Therefore, to consider the nearly optimal solutions, it is convenient to obtain a reduced set, putting the focus on the potentially useful alternatives. These solutions are the alternatives that are close to the optimal solutions in objective space, but which differ significantly in the decision space. To characterize this set, it is essential to simultaneously analyze the decision and objective spaces. One of the crucial points in an evolutionary multi-objective optimization algorithm is the archiving strategy. This is in charge of keeping the solution set, called the archive, updated during the optimization process. The motivation of this work is to analyze the three existing archiving strategies proposed in the literature (ArchiveUpdatePQ,ϵDxy, Archive_nevMOGA, and targetSelect) that aim to characterize the potentially useful solutions. The archivers are evaluated on two benchmarks and in a real engineering example. The contribution clearly shows the main differences between the three archivers. This analysis is useful for the design of evolutionary algorithms that consider nearly optimal solutions.

A synthetic biology approach for the design of genetic algorithms with bacterial agents

ABSTRACT Bacteria have been a source of inspiration for the design of evolutionary algorithms. At the beginning of the twentieth century synthetic biology was born, a discipline whose goal is the design of biological systems that do not exist in nature, for example, programmable synthetic bacteria. In this paper, we introduce as a novelty the designing of evolutionary algorithms where all the steps are conducted by synthetic bacteria. To this end, we designed a genetic algorithm, which we have named BAGA, illustrating its utility solving simple instances of optimisation problems such as function optimisation, 0/1 knapsack problem, Hamiltonian path problem. The results obtained open the possibility of conceiving evolutionary algorithms inspired by principles, mechanisms and genetic circuits from synthetic biology. In summary, we can conclude that synthetic biology is a source of inspiration either for the design of evolutionary algorithms or for some of their steps, as shown by the results obtained in our simulation experiments.

A parallel programming approach to the solution of the location-inventory and multi-echelon routing problem in the humanitarian supply chain

Disasters around the world are becoming more frequent, diverse, complex and extremely challenging, causing millions of casualties and affecting both human development and available resources. Consequently, the present study addresses a multi-objective location, inventory and multi-scale routing (2E-LIRP) problem, which supports comprehensive decision making, so that the logistics network designer and manager can obtain adequate strategic planning in the face of uncertainty and the negative impact that an adverse event can generate. Moreover, the problem is formulated as an integer linear programming model, having as main objectives to minimize private logistics costs and maximize the welfare of the affected areas, considering dynamic demand, multiple products and heterogeneous fleet. Due to the computational complexity associated with the model, new solution approaches are proposed, based on the design of evolutionary metaheuristic algorithms; the first one, known as Non-dominated Sorting Genetic Algorithm version II (NSGA-II), the second one, Strength Pareto Evolutionary Algorithm version II (SPEA-II) and the third one, called Genetic Algorithm (GA), programmed in parallel and executed individually under a cooperative environment.

Guest Editorial Special Issue on Theoretical Foundations of Evolutionary Computation

It is our pleasure to introduce this special issue on the recent advances in the theoretical foundations of evolutionary computation (EC). While in the early days of this field, theoretical analyses inevitably focused on simplified models of evolutionary algorithms (EAs), the continuous progress made in the development of suitable mathematical techniques for the analysis now allows to derive proven statements regarding the performance of off-the-shelf metaheuristics, such as standard generational and steady-state genetic algorithms with no algorithmic simplifications. Comparisons between the performance of a given EA with the best possible one can also be made nowadays, allowing to assess whether a given algorithm may be improved upon or whether its performance is optimal for a given class of problems. Such understanding often provides insights for the design of new EAs which provably have better performance for given problems. We are glad that examples of results of this kind are present within this special issue. A total of 27 papers were submitted which were the subject of at least three independent reviews, and six manuscripts of the highest quality were selected for publication in the special issue. In the following, we provide a brief summary of these manuscripts.