Strengthened Initialization of Adaptive Cross-Generation Differential Evolution

Adaptive Cross-Generation Differential Evolution(ACGDE)is a recently-introduced algorithm for solving multiobjective problems with remarkable performance compared to other evolutionary algorithms(EAs).However,its convergence and diversity are not satisfactory compared with the latest algorithms.In order to adapt to the current environment,ACGDE requires improvements in many aspects,such as its initialization and mutant operator.In this paper,an enhanced version is proposed,namely SIACGDE.It incorporates a strengthened initialization strategy and optimized parameters in contrast to its predecessor.These improvements make the direction of crossgeneration mutation more clearly and the ability of searching more efficiently.The experiments show that the new algorithm has better diversity and improves convergence to a certain extent.At the same time,SIACGDE outperforms other state-of-the-art algorithms on four metrics of 24 test problems.

Hybridization of Differential Evolution and Adaptive-Network-Based Fuzzy Inference Systemin Estimation of Compression Coefficient of Plastic Clay Soil

One of the important geotechnical parameters required for designing of the civil engineering structure is the compressibility of the soil.In this study,the main purpose is to develop a novel hybrid Machine Learning(ML)model(ANFIS-DE),which used Differential Evolution(DE)algorithm to optimize the predictive capability of Adaptive-Network-based Fuzzy Inference System(ANFIS),for estimating soil Compression coefficient(Cc)from other geotechnical parameters namelyWater Content,Void Ratio,SpecificGravity,Liquid Limit,Plastic Limit,Clay content and Depth of Soil Samples.Validation of the predictive capability of the novel model was carried out using statistical indices:Root Mean Square Error(RMSE),Mean Absolute Error(MAE),and Correlation Coefficient(R).In addition,two popular ML models namely Reduced Error Pruning Trees(REPTree)and Decision Stump(Dstump)were used for comparison.Results showed that the performance of the novel model ANFIS-DE is the best(R=0.825,MAE=0.064 and RMSE=0.094)in comparison to other models such as REPTree(R=0.7802,MAE=0.068 and RMSE=0.0988)andDstump(R=0.7325,MAE=0.0785 and RMSE=0.1036).Therefore,the ANFIS-DE model can be used as a promising tool for the correct and quick estimation of the soil Cc,which can be employed in the design and construction of civil engineering structures.

An Efficient Differential Evolution for Truss Sizing Optimization Using AdaBoost Classifier

Design constraints verification is the most computationally expensive task in evolutionary structural optimization due to a large number of structural analyses thatmust be conducted.Building a surrogatemodel to approximate the behavior of structures instead of the exact structural analyses is a possible solution to tackle this problem.However,most existing surrogate models have been designed based on regression techniques.This paper proposes a novel method,called CaDE,which adopts a machine learning classification technique for enhancing the performance of the Differential Evolution(DE)optimization.The proposed method is separated into two stages.During the first optimization stage,the original DE is implemented as usual,but all individuals produced in this phase are stored as inputs of the training data.Based on design constraints verification,these individuals are labeled as“safe”or“unsafe”and their labels are saved as outputs of the training data.When collecting enough data,an AdaBoost model is trained to evaluate the safety state of structures.This model is then used in the second stage to preliminarily assess new individuals,and unpromising ones are rejected without checking design constraints.This method reduces unnecessary structural analyses,thereby shortens the optimization process.Five benchmark truss sizing optimization problems are solved using the proposed method to demonstrate its effectiveness.The obtained results show that the CaDE finds good optimal designs with less structural analyses in comparison with the original DE and four other DE variants.The reduction rate of five examples ranges from 18 to over 50%.Moreover,the proposed method is applied to a real-size transmission tower design problem to exhibit its applicability in practice.

一种差分演化Q表的改进Q-Learning方法

针对Q-Learning算法在路径搜索应用中的盲目性而导致收敛速度慢、回报效率低的问题,文中提出了一种差分演化Q表的改进Q-Learning方法(DE-Q-Learning)。改进算法利用差分演化算法的全局搜索优势,将由Q表个体组成的演化种群通过变异、交叉和选择操作选择出较好的初始Q表,以此提升Q-Learning前期回报与探索能力。文中在OpenAI的Gym环境中验证了DE-Q-Learning方法的有效性,并进一步在复杂迷宫环境和强化学习环境Pacman中实验了其在复杂路径搜索和动态避障问题上的性能。实验结果表明,DE-Q-Learning在Pacman环境中相比于改进算法Double-Q-Learning和SA-Q-Learning不仅在历史回报方面具有明显优势,而且收敛速度分别提升了42.16%和15.88%,这表明DE-Q-Learning能够显著提高历史累积回报和算法的收敛速度。

Statistical learning makes the hybridization of particle swarm and differential evolution more efficient-A novel hybrid optimizer

This brief paper reports a hybrid algorithm we developed recently to solve the global optimization problems of multimodal functions, by combining the advantages of two powerful population-based metaheuristics differential evolution （DE） and particle swarm optimization （PSO）. In the hybrid denoted by DEPSO, each individual in one generation chooses its evolution method, DE or PSO, in a statistical learning way. The choice depends on the relative success ratio of the two methods in a previous learning period. The proposed DEPSO is compared with its PSO and DE parents, two advanced DE variants one of which is suggested by the originators of DE, two advanced PSO variants one of which is acknowledged as a recent standard by PSO community, and also a previous DEPSO. Benchmark tests demonstrate that the DEPSO is more competent for the global optimization of multimodal functions due to its high optimization quality.

Quantum learning control using differential evolution with equally-mixed strategies

Learning control has been recognized as a powerful approach in quantum information technology. In this paper, we extend the application of differential evolution （DE） to design optimal control for various quantum systems. Various DE methods are introduced and analyzed, and EMSDE featuring in equally mixed strategies is employed for quantum control. Two classes of quantum control problems, including control of four-level open quantum ensembles and quantum superconducting systems, are investigated to demonstrate the performance of EMSDE for learning control of quantum systems. Numerical results verify the effectiveness of the FMSDE method for various quantum systems and show the potential for complex quantum control problems.

Solving Nonlinear Equations Systems with an Enhanced Reinforcement Learning Based Differential Evolution

Nonlinear equations systems(NESs)arise in a wide range of domains.Solving NESs requires the algorithm to locate multiple roots simultaneously.To deal with NESs efficiently,this study presents an enhanced reinforcement learning based differential evolution with the following major characteristics:(1)the design of state function uses the information on the fitness alternation action;(2)different neighborhood sizes and mutation strategies are combined as optional actions;and(3)the unbalanced assignment method is adopted to change the reward value to select the optimal actions.To evaluate the performance of our approach,30 NESs test problems and 18 test instances with different features are selected as the test suite.The experimental results indicate that the proposed approach can improve the performance in solving NESs,and outperform several state-of-the-art methods.

Differential Evolution with Level-Based Learning Mechanism

To address complex single objective global optimization problems,a new Level-Based Learning Differential Evolution(LBLDE)is developed in this study.In this approach,the whole population is sorted from the best to the worst at the beginning of each generation.Then,the population is partitioned into multiple levels,and different levels are used to exert different functions.In each level,a control parameter is used to select excellent exemplars from upper levels for learning.In this case,the poorer individuals can choose more learning exemplars to improve their exploration ability,and excellent individuals can directly learn from the several best individuals to improve the quality of solutions.To accelerate the convergence speed,a difference vector selection method based on the level is developed.Furthermore,specific crossover rates are assigned to individuals at the lowest level to guarantee that the population can continue to update during the later evolutionary process.A comprehensive experiment is organized and conducted to obtain a deep insight into LBLDE and demonstrates the superiority of LBLDE in comparison with seven peer DE variants.

Adaptive Dimensional Learning with a Tolerance Framework for the Differential Evolution Algorithm

The Differential Evolution(DE)algorithm,which is an efficient optimization algorithm,has been used to solve various optimization problems.In this paper,adaptive dimensional learning with a tolerance framework for DE is proposed.The population is divided into an elite subpopulation,an ordinary subpopulation,and an inferior subpopulation according to the fitness values.The ordinary and elite subpopulations are used to maintain the current evolution state and to guide the evolution direction of the population,respectively.The inferior subpopulation learns from the elite subpopulation through the dimensional learning strategy.If the global optimum is not improved in a specified number of iterations,a tolerance mechanism is applied.Under the tolerance mechanism,the inferior and elite subpopulations implement the restart strategy and the reverse dimensional learning strategy,respectively.In addition,the individual status and algorithm status are used to adaptively adjust the control parameters.To evaluate the performance of the proposed algorithm,six state-of-the-art DE algorithm variants are compared on the benchmark functions.The results of the simulation show that the proposed algorithm outperforms other variant algorithms regarding function convergence rate and solution accuracy.

Opposition-based differential evolution for hydrothermal power system

This paper presents opposition-based differential evolution to determine the optimal hourly schedule of power generation in a hydrothermal system.Differential evolution(DE)is a population-based stochastic parallel search evolutionary algorithm.Opposition-based differential evolution has been used here to improve the effectiveness and quality of the solution.The proposed opposition-based differential evolution(ODE)employs opposition-based learning(OBL)for population initialization and also for generation jumping.The effectiveness of the proposed method has been verified on two test problems,two fixed head hydrothermal test systems and three hydrothermal multi-reservoir cascaded hydroelectric test systems having prohibited operating zones and thermal units with valve point loading.The results of the proposed approach are compared with those obtained by other evolutionary methods.It is found that the proposed opposition-based differential evolution based approach is able to provide better solution.

电推进GEO卫星的改进粒子群轨道保持优化设计

针对地球同步轨道(GEO)卫星轨道保持问题,提出了一种基于改进粒子群算法(PSO)的序列电推力轨道保持方法。首先,建立了GEO卫星高精度非线性轨道动力学模型和序列电推力模型。然后,设计了GEO卫星相对轨道保持策略,建立了以燃料消耗为性能指标的序列电推力轨道保持问题优化模型并进行了离散化。接着,通过引入差分进化算法和维度学习策略对粒子群优化算法进行了适应性改进,同时对推力大小和作用时间进行寻优计算。最后,通过数值仿真对所提出的改进粒子群优化算法进行了对比校验。结果表明,该方法在完成GEO卫星轨道保持任务的同时具备燃料消耗低和收敛速度快等优点。

一种用于变压器故障诊断的贝叶斯网络优化方法

针对变压器故障诊断效率低的问题,文中将油中溶解气体分析与人工智能方法相结合,提出了一种改进蝗虫优化算法优化贝叶斯网络的变压器故障诊断方法。利用差分进化算法和与模拟退火算法对蝗虫算法进行改进,提高了算法的优化能力。将改进蝗虫算法应用于贝叶斯网络结构来学习构建变压器故障诊断模型,利用所提方法对变压器进行故障诊断。实验结果表明,该方法诊断正确率达到了92.7%,与其他算法所构建的诊断模型相比具有更高的故障诊断准确率。

基于非正交多址的多无人机协同计算与任务卸载策略

人群密集的应用场景下存在计算资源需求量大、无人机计算任务分配不均等计算需求问题。针对该类场景,论文提出一种基于非正交多址的多无人机辅助的移动边缘计算任务卸载方案。首先,利用非正交多址和串行干扰删除技术提升用户的传输速率,并利用多无人机的相互协作防止任务分配不均。在符合用户设备能耗、计算资源的前提下,通过联合优化无人机部署位置和卸载策略,构建一个使系统能耗最小化的优化问题。并将该优化问题分解为两个子问题,利用深度强化学习中的DDQN网络求出用户的卸载决策,利用差分进化算法确定此卸载决策下的无人机部署,然后交替迭代两种方法得到问题的优化解。仿真结果表明相较于时分多址技术,非正交多址有效地降低了用户任务的传输时延。相较于DQN网络、贪婪算法,论文所提卸载决策算法有效降低了系统总能耗。

基于差分进化算法的瞬变电磁一维反演

实际采集的瞬变电磁数据包含电磁感应和激发极化效应,如何准确提取电阻率和极化率信息是电性源瞬变电磁数据处理的关键。首先,基于Cole⁃Cole复电阻率模型实现有限长电性源瞬变电磁法一维正演,在此基础上提出一种基于差分进化算法的电性源瞬变电磁一维反演方法。然后,在传统差分进化算法的基础上引入反向学习策略及控制参数自适应调节,加快反演的收敛速度,同时在目标函数中引入约束条件,构成最小构造反演,降低反演的多解性。最后,基于典型的三层地电模型和复杂多层模型进行理论模型测试,反演结果可有效恢复模型的电阻率和极化率。利用实测资料进行反演,反演得到的电阻率与OCCAM反演电阻率基本一致。在此电阻率约束的基础上,进一步反演得到极化率信息。反演结果准确地提取了实测数据中的电阻率信息,得到了地下介质的极化率分布,证明了算法的准确性和适用性。

基于改进差分进化算法的动态防空资源分配优化

面对动态防空资源分配问题中存在的空袭目标突现和雷达、发射车等资源受干扰现象,在综合考虑雷达、发射车和导弹等武器装备性能的基础上,基于目标集、资源集建立了最小化目标总拦截价值与生存概率的混合整数决策模型。提出了一种新的改进差分进化算法进行求解,采用反向学习策略生成初始解,确保初始种群的质量,设计了一种快速修复与重构的启发式规则作用于多阶段,以提升算法的搜索能力。仿真实验验证了该算法具有求解时间和求解精度上的优越性。该研究能使武器系统在动态事件的随机影响下,保持高效的作战能力和决策效果。

Q学习差分进化算法求解热电动态经济排放调度

热电联产动态经济排放调度同时考虑了燃料成本花费和污染气体排放两个目标值,且下一时间段的热电产量受当前时间段热电产量的影响,这是近年来电力系统运行中的一个重要问题。文中提出一种基于Q学习强化多目标差分进化(Q Learning Multi-Objective Differential Evolution,QLMODE)算法,以此求解热电联产动态经济排放调度(Combined Heat and Power Dynamic Economic Emission Dispatch,CHPDEED)问题。在QLMODE中,采用Q学习技术调整算法的比例因子参数,即在迭代过程中利用子代解和父代解之间的支配关系确定动作奖励和惩罚,并通过Q学习调整参数值,以获得最适合环境模型的算法参数。文中将所提QLMODE用于求解11机组和33机组的热电联产动态经济排放调度问题。仿真结果表明,与4种成熟的多目标优化算法相比,QLMODE算法燃料成本最小,污染气体排放最少,收敛性和多样性指标优于其他4种算法,且QLMODE在两组问题上都获得了更好的Pareto最优前沿。

一种融合反向学习机制与差分进化策略的蛇优化算法

蛇优化(snake optimizer,SO)算法存在前期收敛速度慢和易陷入局部最优的问题,为此提出一种融合反向学习机制与差分进化策略的改进蛇优化(improved snake optimizer,ISO)算法。反向学习机制可提高种群质量,以提升算法寻优速度;差分进化策略有助于算法精准寻优,降低算法陷入局部最优的几率。在10个基准测试函数上的实验结果表明,ISO算法拥有更高的寻优精度和更快的收敛速率。将其应用于支持向量机(support vector machine,SVM)的参数选取中,进一步验证了ISO算法的有效性。

Learning to select the recombination operator for derivative-free optimization

Extensive studies on selecting recombination operators adaptively,namely,adaptive operator selection(AOS),during the search process of an evolutionary algorithm(EA),have shown that AOS is promising for improving EA's performance.A variety of heuristic mechanisms for AOS have been proposed in recent decades,which usually contain two main components:the feature extraction and the policy setting.The feature extraction refers to as extracting relevant features from the information collected during the search process.The policy setting means to set a strategy(or policy)on how to select an operator from a pool of operators based on the extracted feature.Both components are designed by hand in existing studies,which may not be efficient for adapting optimization problems.In this paper,a generalized framework is proposed for learning the components of AOS for one of the main streams of EAs,namely,differential evolution(DE).In the framework,the feature extraction is parameterized as a deep neural network(DNN),while a Dirichlet distribution is considered to be the policy.A reinforcement learning method,named policy gradient,is used to train the DNN.As case studies,the proposed framework is applied to two DEs including the classic DE and a recently-proposed DE,which result in two new algorithms named PG-DE and PG-MPEDE,respectively.Experiments on the Congress of Evolutionary Computation(CEC)2018 test suite show that the proposed new algorithms perform significantly better than their counterparts.Finally,we prove theoretically that the considered classic methods are the special cases of the proposed framework.

基于PCA-SaDE-ELM优化算法的煤层底板破坏深度预测及工程应用

基于煤层底板破坏深度实测结果统计分析,通过优化数据样本空间,引入自适应差分进化改进的极限学习机算法,构建了煤层底板破坏深度预测模型,与实测结果对比分析验证,并应用于云驾岭煤矿9^(#)煤层底板破坏深度预测。结果表明:模型预测的最大绝对误差不超过0.7 m,相比现有其他预测模型,该模型预测精度提高约70%;云驾岭煤矿19101、19103和19105这3个典型工作面的破坏深度分别为10.80、10.94、11.34 m,介于规范方法和滑移场理论预测结果之间,进一步反映了模型的可靠性;建议对9#煤层底板加固改造后再进行回采。相关研究成果可为我国煤层底板破坏风险管理和煤炭资源的优化回采布置提供一定的理论支撑。

基于最优解区间预筛选的代理模型辅助天线设计优化算法

针对天线优化中全波仿真计算耗时过多的问题,文中提出了一种基于数据约束的代理模型辅助进化算法(SAADC)以实现天线优化设计中效率的提升。首先采用增强随机型差分进化算法,以保证数据生成的随机性与多样性。进一步通过高斯代理模型对仿真结果进行预测,并利用最优解区间预筛选方法舍弃预测结果中较差的个体,以实现算法收敛速度的提升。最终利用所提出的SAADC,对三个不同拓扑结构E型贴片天线的带宽与增益进行了优化设计与结果分析。结果表明,所提出的算法比现有的代理模型优化算法具有更快的优化速度与更佳的优化结果,可满足天线结构高效优化的实际需求。