Dynamic Evolutionary Algorithm Research Articles

Automatic deep sparse clustering with a dynamic population-based evolutionary algorithm using reinforcement learning and transfer learning

Clustering data effectively remains a significant challenge in machine learning, particularly when the optimal number of clusters is unknown. Traditional deep clustering methods often struggle with balancing local and global search, leading to premature convergence and inefficiency. To address these issues, we introduce ADSC-DPE-RT (Automatic Deep Sparse Clustering with a Dynamic Population-based Evolutionary Algorithm using Reinforcement Learning and Transfer Learning), a novel deep clustering approach. ADSC-DPE-RT builds on Multi-Trial Vector-based Differential Evolution (MTDE), an algorithm that integrates sparse auto-encoding and manifold learning to enable automatic clustering without prior knowledge of cluster count. However, MTDE's fixed population size can lead to either prolonged computation or premature convergence. Our approach introduces a dynamic population generation technique guided by Reinforcement Learning (RL) and Markov Decision Process (MDP) principles. This allows for flexible adjustment of population size, preventing premature convergence and reducing computation time. Additionally, we incorporate Generative Adversarial Networks (GANs) to facilitate dynamic knowledge transfer between MTDE strategies, enhancing diversity and accelerating convergence towards the global optimum. This is the first work to address the dynamic population issue in deep clustering through RL, combined with Transfer Learning to optimize evolutionary algorithms. Our results demonstrate significant improvements in clustering performance, positioning ADSC-DPE-RT as a competitive alternative to state-of-the-art deep clustering methods.

A Dynamic Tasking-Based Evolutionary Algorithm for Bi-Objective Feature Selection

Feature selection in classification is a complex optimization problem that cannot be solved in polynomial time. Bi-objective feature selection, aiming to minimize both selected features and classification errors, is challenging due to the conflict between objectives, while one of the most effective ways to tackle this is to use multi-objective evolutionary algorithms. However, very few of these have ever reflected an evolutionary multi-tasking framework, despite the implicit parallelism offered by the population-based search characteristic. In this paper, a dynamic multi-tasking-based multi-objective evolutionary algorithm (termed DTEA) is proposed for handling bi-objective feature selection in classification, which is not only suitable for datasets with relatively lower dimensionality of features, but is also suitable for datasets with relatively higher dimensionality of features. The role and influence of multi-tasking on multi-objective evolutionary feature selection were studied, and a dynamic tasking mechanism is proposed to self-adaptively assign multiple evolutionary search tasks by intermittently analyzing the population behaviors. The efficacy of DTEA is tested on 20 classification datasets and compared with seven state-of-the-art evolutionary algorithms. A component contribution analysis was also conducted by comparing DTEA with its three variants. The empirical results show that the dynamic-tasking mechanism works efficiently and enables DTEA to outperform other algorithms on most datasets in terms of both optimization and classification.

Traveling of multiple salesmen to dynamically changing locations for satisfying multiple goals

Polymer grade scheduling, maritime surveillance, e-food delivery, e-commerce, and military tactics necessitate multiple agents (e.g., extruders, speed boats, salesmen) capable of visiting (or completing) dynamically changing locations (or tasks) in minimum time and distance. This study proposes a novel methodology based on clustering and local heuristic-based evolutionary algorithms to address the dynamic traveling salesman problem (TSP) and the dynamic multi-salesman problem with multiple objectives. The proposed algorithm is evaluated on 11 benchmark TSP problems and large-scale problems with up to 10,000 instances. The results show the superior performance of the proposed methodology called the dynamic two-stage evolutionary algorithm as compared to the dynamic hybrid local search evolutionary algorithm. Furthermore, the algorithm's applicability is illustrated through various scenarios involving up to four salesmen and three objectives with dynamically changing locations. To demonstrate real-world relevance, a maritime surveillance problem employing a helideck monitoring system is solved, wherein the objective is to minimize the patrolling route while visiting faulty vessels that threaten marine vessels. This study provides a general framework of TSP which finds application in several sectors, including planning and scheduling in chemical and manufacturing industries, the defense sector, and the e-commerce sector. Finally, the results showcase the effectiveness of the proposed methodology in solving the dynamic multiobjective, and multiple salesmen problem, which represents a more generalized version of the TSP.

A dynamic tri-population multi-objective evolutionary algorithm for constrained multi-objective optimization problems

A dynamic tri-population multi-objective evolutionary algorithm for constrained multi-objective optimization problems

An Effective Dynamic Constrained Two-Archive Evolutionary Algorithm for Cooperative Search-Track Mission Planning by UAV Swarms in Air Intelligent Transportation

An Effective Dynamic Constrained Two-Archive Evolutionary Algorithm for Cooperative Search-Track Mission Planning by UAV Swarms in Air Intelligent Transportation

Evolving masked low-rank transformer for long text understanding

Long sequence text processing is time-consuming owing to the ultra-large-scale self-attention computing. Recent advances demonstrate the attention in transformer can be accelerated by redundancy removal, and there are various sparse variants for attention in large sequences are proposed, which leads to state-of-the-art performance on language and vision task. Low-rank method achieve outstanding success in the field of efficient transformer. The dynamic token sparsification is efficiently time-saving and cost-saving, which can be easily extended to prune redundant spans and to yield semantic features. Evolutionary algorithm is attractive for selecting hyperparameter which is of significant importance in effectiveness. Motivated by these works, we propose an efficient transformers model, termed EMLT, to alleviate time and cost without sacrificing the accuracy. EMLT effectively combines strengths of Low-rank transformers, dynamic token sparsification and evolutionary algorithm to ulteriorly cut redundant token and meanwhile maintains the original precision, which can achieve a linear memory and time complexity. We compress transformer in three stages. Firstly, sliding window is validated as local attention to capture fine-grained dependency semantics. After that, low-rank approximation of attention matrix is applied as global attention to store long-range dependency semantics, and aggregated with local attention. On this basis, we consistently prune redundant token in accordance with importance score to further sparse the attention operation. Finally, Evolutionary algorithm is utilized to optimize the hyper-parameters of every layer. The results of comprehensive experiments and analysis show that our method can rival others on accuracy, and outperforms others on efficiency by a large margin in terms of the computational complexity.

A decomposition and dynamic niching distance-based dual elite subpopulation evolutionary algorithm for multimodal multiobjective optimization

Solving multimodal multiobjective optimization problems(MMOPs) require obtaining multiple Pareto optimal solution sets (PSs). In recent years, researchers have proposed some multimodal multiobjective evolutionary algorithms (MMEAs), however, many existing MMEAs perform unsatisfactorily in dealing with complicated MMOPs, and even fail to find all Pareto solution sets in a stable manner. Many existing algorithms rely on globally optimal individuals to guide population evolution, which leads to the population rapidly converging to a few easy-to-find Pareto solution sets (PSs) and losing hard-to-find PSs. MMEAs that cannot consistently find all PSs are difficult to handle the increasing complexity of MMOPs. In order to address these problems, this research proposes a decomposition and dynamic niching distance-based dual elite subpopulation evolutionary algorithm, named MMEA-DES. In the proposed algorithm, decomposition-based methods are employed to maintain the diversity of the objective space of MMOPs, and a dynamic niching distance (DND) is proposed to calculate the crowding of the decision space. In addition, the dual elite subpopulations (DES) based on decomposition and DND are implemented to support the algorithm in obtaining well-distributed solutions in both decision space and objective space. To validate the performance of the proposed algorithm, a comparison with state-of-the-art algorithms on 23 test problems shows that the algorithm proposed in this research can stably solve various types of complicated MMOPs.

A dynamic constrained multiobjective evolutionary algorithm based on decision variable classification

In dynamic constrained multiobjective optimization problems (DCMOPs), dynamics may arise from time-varying objective functions or/and constraints. To solve these problems, maintaining a good balance among feasibility, convergence and diversity of a population under dynamic environments is a critical challenge. Although appropriately utilizing the characteristic of decision variables can promote algorithms to better track the Pareto optima under dynamic environments, their sensitivity to constraints is neglected. Therefore, a dynamic constrained multiobjective evolutionary algorithm based on decision variable classification (DC-MOEA-DVC) is proposed. Under each environment, decision variables are classified into four types in terms of their influence on convergence, distribution, and constraint violation. Based on them, a new offspring generation method is developed, decision variables with different characteristics are rationally combined to generate offspring, with the purpose of accelerating the convergence of the population. Once an environmental change appears, a hybrid strategy consisting of four change response techniques is introduced for the corresponding types of decision variables, producing a new initial population. The experimental results show that DC-MOEA-DVC is superior to the other five state-of-the-art algorithms.

A dynamic core evolutionary clustering algorithm based on saturated memory

Because the number of clustering cores needs to be set before implementing the K-means algorithm, this type of algorithm often fails in applications with increasing data and changing distribution characteristics. This paper proposes an evolutionary algorithm DCC, which can dynamically adjust the number of clustering cores with data change. DCC algorithm uses the Gaussian function as the activation function of each core. Each clustering core can adjust its center vector and coverage based on the response to the input data and its memory state to better fit the sample clusters in the space. The DCC algorithm model can evolve from 0. After each new sample is added, the winning dynamic core can be adjusted or split by competitive learning, so that the number of clustering cores of the algorithm always maintains a better adaptation relationship with the existing data. Furthermore, because its clustering core can split, it can subdivide the densely distributed data clusters. Finally, detailed experimental results show that the evolutionary clustering algorithm DCC based on the dynamic core method has excellent clustering performance and strong robustness.

A bidirectional dynamic grouping multi-objective evolutionary algorithm for feature selection on high-dimensional classification

As a key preprocessing step in classification, feature selection involves two conflicting objectives: maximizing the classification accuracy and minimizing the number of selected features. Therefore, multi-objective optimization is widely used in feature selection due to its excellent trade-off between the convergence of two objectives. However, most existing multi-objective feature selection methods still face the issues of the “curse of dimensionality” and high computational costs, especially when the search space is large. To solve the above issues, this paper proposes a bidirectional dynamic grouping multi-objective evolutionary approach for high-dimensional feature selection, referred to as BDGMOEA. This approach transforms a high-dimensional feature selection problem into a feature selection task with a smaller search space by the idea of feature grouping, in which one bit of an individual represents a group of features. Specifically, a grouping search strategy is developed to divide the features into different quadrants according to the importance of the features obtained by different evaluation techniques. Then, the features in each quadrant are grouped by sector. This strategy can effectively narrow the search space and quickly locate promising feature regions. In addition, a bidirectional dynamic adjustment mechanism is presented by considering the evolutionary state of the population, and it can be used to explore each feature in more detail and comprehensively to prevent good features from being ignored in unselected groups. The experimental results demonstrate that the proposed BDGMOEA method performs the best in most cases, indicating that BDGMOEA not only achieves better classification performance but also reduces the training time.

Dynamic matrix-based evolutionary algorithm for large-scale sparse multiobjective optimization problems

Dynamic matrix-based evolutionary algorithm for large-scale sparse multiobjective optimization problems

A multiobjective dynamic rebalancing evolutionary algorithm for free-floating bike sharing

The big challenge of free-floating bike sharing system is to rebalance the supply and demand in different places. Most studies are static rebalancing that moves bikes at night. Dynamic rebalancing considers the scheduling of transport vehicles during the day. It is more effective and difficult to solve. In this paper, a multiobjective dynamic rebalancing problem is solved. The two optimization objectives are considered: minimizing the total distance of rebalancing vehicle and minimizing the total demand unmet. A mathematic model is built to formulate the problem. A new algorithm SFMEA-AP (Sequence Flow Multiobjective Evolutionary Algorithm with Area Protected) that uses the sequence flow encoding is proposed. An novel area protected strategy that includes a non-dominated sorting with similarity comparison is proposed. It simultaneously considers the fairness and competitiveness in solution space. Six neighborhood search operators are designed. They are used as variable neighborhood search strategy. Computational experiments show that SFMEA-AP can obtain better solution quality than SFMEA-VNS, QMOEA, NSGA-II, ILS and LSMOVRPTW. The IGD values of non-dominated solutions of SFMEA-AP are better than those of SFMEA-VNS, QMOEA, NSGA-II, ILS and LSMOVRPTW. In addition, the non-dominated solution set of SFMEA-AP have good diversity.

A dynamic dual-population co-evolution multi-objective evolutionary algorithm for constrained multi-objective optimization problems

Many multi-objective evolutionary algorithms are proposed to handle constrained multi-objective optimization problems. Nevertheless, they often fail to appropriately balance feasibility, convergence and diversity of the population. This paper proposed a dynamic dual-population co-evolution multi-objective evolutionary algorithm (DDCMEA) to solve this issue. In DDCMEA, a dynamic dual-population co-evolution strategy is employed to balance the convergence and the feasibility by dynamically adjusting the offspring number of the two populations. In the early stage of evolution, the algorithm mainly focuses on the convergence and more offspring of the first population are generated. In the late stage of evolution, the algorithm mainly focuses on the feasibility and more offspring of the second population are generated. Finally, feasible solutions with good convergence could be obtained. To further enhance the diversity of the offspring and obtain feasible solutions with a wide spread of distribution, the evolution operators of the genetic algorithm and the differential evolution are chosen as the search engines for the first population and the second population, respectively. The performance of DDCMEA is further tested through thirty-one bench-mark test problems and two real-world problems in comparison with other five state-of-the-art algorithms. The results show the proposed algorithm DDCMEA achieves competitive performance when handling constrained multi-objective optimization problems.

An evolutionary algorithm based on dynamic sparse grouping for sparse large scale multiobjective optimization

Sparse large scale multiobjective optimization problems (sparse LSMOPs) contain numerous decision variables, and their Pareto optimal solutions' decision variables are very sparse (i.e., the majority of these solutions' decision variables are zero-valued). This poses grand challenges to an algorithm in converging to the Pareto set. Numerous evolutionary algorithms (EAs) tailored for sparse LSMOPs have been proposed in recent years. However, the final population generated by these EAs is not sparse enough because the location of the nonzero decision variables is difficult to locate accurately and there is insufficient interaction between the nonzero decision variables' locating process and the nonzero decision variables' optimizing process. To address this issue, we propose a dynamic sparse grouping evolutionary algorithm (DSGEA) that dynamically groups decision variables in the population that have a comparable amount of nonzero decision variables. Improved evolutionary operators are introduced to optimize the decision variables in groups. As a result, the population obtained by DSGEA can stably evolve towards the sparser Pareto optimal that has a precise location of nonzero decision variables. The proposed algorithm outperforms existing up-to-date EAs for sparse LSMOPs in experiments on three real-world problems and eight benchmark problems.

Scheduling and Process Optimization for Blockchain-Enabled Cloud Manufacturing Using Dynamic Selection Evolutionary Algorithm

The blockchain-enabled cloud manufacturing is an emerging service-oriented paradigm, and the scheduling and process optimization for blockchain-enabled cloud manufacturing (SPO-BCMfg) are crucial to achieving the service-oriented goal. The blockchain-enabled cloud manufacturing paradigm improves the collaboration capabilities and information security over the ordinary cloud manufacturing while incorporating distributed storage, consensus mechanism, and cloud-edge collaboration. The above characteristics make SPO-BCMfg a multiobjective scheduling optimization problem. This article establishes the multiobjective SPO-BCMfg model based on a dynamic selection evolutionary algorithm to address the problem. First, we carry out the architecture and the modeling of the blockchain cloud manufacturing system. Then, a novel dynamic selection evolutionary algorithm is proposed, which is used to schedule and optimize the model for the process. In the stage of evolution, the algorithm uses a diversity-based population partitioning technique that utilizes the dynamic distance to realize the selection of elite solutions. The method was experimented on the SPO-BCMfg problem facing five and eight objectives. The experimental results show that the algorithm has a strong processing capacity in terms of convergence and diversity compared with the other advanced evolutionary algorithms.

A Two-State Dynamic Decomposition-Based Evolutionary Algorithm for Handling Many-Objective Optimization Problems

Decomposition-based many-objective evolutionary algorithms (D-MaOEAs) are brilliant at keeping population diversity for predefined reference vectors or points. However, studies indicate that the performance of an D-MaOEA strongly depends on the similarity between the shape of the reference vectors (points) and that of the PF (a set of Pareto-optimal solutions symbolizing balance among objectives of many-objective optimization problems) of the many-objective problem (MaOP). Generally, MaOPs with expected PFs are not realistic. Consequently, the inevitable weak similarity results in many inactive subspaces, creating huge difficulties for maintaining diversity. To address these issues, we propose a two-state method to judge the decomposition status according to the number of inactive reference vectors. Then, two novel reference vector adjustment strategies, set as parts of the environmental selection approach, are tailored for the two states to delete inactive reference vectors and add new active reference vectors, respectively, in order to ensure that the reference vectors are as close as possible to the PF of the optimization problem. Based on the above strategies and an efficient convergence performance indicator, an active reference vector-based two-state dynamic decomposition-base MaOEA, referred to as ART-DMaOEA, is developed in this paper. Extensive experiments were conducted on ART-DMaOEA and five state-of-the-art MaOEAs on MaF1-MaF9 and WFG1-WFG9, and the comparative results show that ART-DMaOEA has the most competitive overall performance.

A Novel Clustering Strategy-Based Sink Path Optimization for Wireless Sensor Network

To optimize the path of mobile sinks in the wireless sensor network (WSN) and use of 5G signals to transmit information, a new clustering strategy is proposed, which not only balances the number of cluster nodes but also shortens the path of mobile sinks. By moving rendezvous points (RPs) to change the cluster position, we first propose a metaheuristics clustering algorithm named the dynamic clustering-based rectangular evolutionary algorithm (DCREA). The algorithm adjusts the clustering structure of the network by limiting the cluster movement range; then, it is optimized in combination with the greedy algorithm. Extensive experimental results show that this method can significantly shorten the path of the mobile sink, increase the efficiency of the mobile sink, and improve the quality of signal transmission while maintaining the balance of the cluster structure. The experimental results show that the DCREA is the most effective and indicate the effectiveness of the algorithm by comparing it with related recent algorithms.

A novel dynamic reference point model for preference-based evolutionary multiobjective optimization

AbstractIn the field of preference-based evolutionary multiobjective optimization, optimization algorithms are required to search for the Pareto optimal solutions preferred by the decision-maker (DM). The reference point is a type of techniques that effectively describe the preferences of DM. So far, the reference point is either static or interactive with the evolutionary process. However, the existing reference point techniques do not cover all application scenarios. A novel case, i.e., the reference point changes over time due to the environment change, has not been considered. This paper focuses on the multiobjective optimization problems with dynamic preferences of the DM. First, we propose a change model of the reference point to simulate the change of the preference by the DM over time. Then, a dynamic preference-based multiobjective evolutionary algorithm framework with a clonal selection algorithm (ĝa-NSCSA) and a genetic algorithm (ĝa-NSGA-II) is designed to solve such kind of optimization problems. In addition, in terms of practical applications, the experiments on the portfolio optimization problems with the dynamic reference point model are tested. Experimental results on the benchmark problems and the practical applications show that ĝa-NSCSA exhibits better performance among the compared optimization algorithms.

A novel dynamic reference point model for preference-based evolutionary multiobjective optimization

In the field of preference-based evolutionary multiobjective optimization, optimization algorithms are required to search for the Pareto optimal solutions preferred by the decision maker (DM). The reference point is a type of techniques that effectively describe the preferences of DM. So far, the reference point is either static or interactive with the evolutionary process. However, the existing reference point techniques do not cover all application scenarios. A novel case, i.e., the reference point changes over time due to the environment change, has not been considered. This paper focuses on the multiobjective optimization problems with dynamic preferences of the DM. First, we propose a change model of the reference point to simulate the change of the preference by the DM over time. Then, a dynamic preference-based multiobjective evolutionary algorithm framework with a clonal selection algorithm (ĝa-NSCSA) and a genetic algorithm (ĝa-NSGA-II) is designed to solve such kind of optimization problems. In addition, in terms of practical applications, the experiments on the portfolio optimization problems with the dynamic reference point model are tested. Experimental results on the benchmark problems and the practical applications show that ĝa-NSCSA exhibits better performance among the compared optimization algorithms.

Optimization-Based Fault Location Algorithm for Series-Compensated Power Transmission Lines

This paper presents an optimization-based fault location algorithm applicable to series-compensated power transmission lines. The presented method determines the fault location minimizing the errors between measured and calculated post-fault voltage and current phasors through a dynamic differential evolutionary optimization algorithm. This fault location solution may be advantageous over existing approaches found in the literature as it can locate any fault regardless of the series compensation’s design, impedance, and positioning. To validate the proposed algorithm, the authors tested it against more than a thousand and six hundred faults on a 500 [kV] series-compensated double-circuit transmission line simulated in ATP/EMTP. The results indicate that the presented method yields accurate fault location estimates, with errors within the 1% magnitude using post-fault steady-state and transient data. In addition, the proposed algorithm is robust even in situations featuring high-resistance faults, errors in the line’s parameters, in the system equivalents connected to and between the line’s terminals, and in phasor extraction, configuring it as an accurate and reliable alternative to locate faults within series-compensated lines.