Scale adaptive fitness evaluation‐based particle swarm optimisation for hyperparameter and architecture optimisation in neural networks and deep learning

Research into automatically searching for an optimal neural network(NN)by optimi-sation algorithms is a significant research topic in deep learning and artificial intelligence.However,this is still challenging due to two issues:Both the hyperparameter and ar-chitecture should be optimised and the optimisation process is computationally expen-sive.To tackle these two issues,this paper focusses on solving the hyperparameter and architecture optimization problem for the NN and proposes a novel light‐weight scale‐adaptive fitness evaluation‐based particle swarm optimisation(SAFE‐PSO)approach.Firstly,the SAFE‐PSO algorithm considers the hyperparameters and architectures together in the optimisation problem and therefore can find their optimal combination for the globally best NN.Secondly,the computational cost can be reduced by using multi‐scale accuracy evaluation methods to evaluate candidates.Thirdly,a stagnation‐based switch strategy is proposed to adaptively switch different evaluation methods to better balance the search performance and computational cost.The SAFE‐PSO algorithm is tested on two widely used datasets:The 10‐category(i.e.,CIFAR10)and the 100−cate-gory(i.e.,CIFAR100).The experimental results show that SAFE‐PSO is very effective and efficient,which can not only find a promising NN automatically but also find a better NN than compared algorithms at the same computational cost.

Adaptive multifactorial particle swarm optimisation

Existing multifactorial particle swarm optimisation(MFPSO)algorithms only explore a relatively narrow area between the inter-task particles.Meanwhile,these algorithms use a fixed inter-task learning probability throughout the evolution process.However,the parameter is problem dependent and can be various at different stages of the evolution.In this work,the authors devise an inter-task learning-based information transferring mechanism to replace the corresponding part in MFPSO.This inter-task learning mechanism transfers the searching step by using a differential term and updates the personal best position by employing an inter-task crossover.By this mean,the particles can explore a broad search space when utilising the additional searching experiences of other tasks.In addition,to enhance the performance on problems with different complementarity,they design a self-adaption strategy to adjust the inter-task learning probability according to the performance feedback.They compared the proposed algorithm with the state-of-the-art algorithms on various benchmark problems.Experimental results demonstrate that the proposed algorithm can transfer inter-task knowledge efficiently and perform well on the problems with different complementarity.

Co-digestion of Waste Coffee and Cocoa Hulls: Modeling of Biogas Production by the Particle Swarm Method

Energy is a crucial material for the development of our economy.Access to sufficient energy remains a major concern for developing countries,particularly those in sub-Saharan Africa.The major challenge lies in access to clean,environmentally friendly,quality and low-cost energy in different households in our municipalities.To cope with this vast energy gap,many households are dependent on fossil fuels.In Cameroon,the consumption of wood for the supply of energy is increasing by 4%per year.Overall,approximately 80%of households in Cameroon depend on woody biomass as the sole main source of energy supply in Cameroon and demand is growing over time.In view of the climatic variations that our countries,particularly Cameroon,undergo through deforestation,the use of wood as a source of energy is expensive and harmful to the environment,hence the urgency of replacing wood with renewable energy.Biogas is one of the most versatile sources of renewable energy.On an industrial scale,it is important to automate the process control.The main objective of the present work is to model the anaerobic digestion of coffee and cocoa hulls using the particle swarm optimisation method.Pretreatment using the organosolv process was done.This resulted in 48%lignin removal and 22%cellulose increase.For the pretreated biomass,the maximum production rate was 21 NmLCH4 per day with a biomethane yield of 90 NmLCH4/gVS.This represents an enhancement of 117%in biomethane yield.A positive flammability test was recorded after the 10th day of retention time.Moreover,the data collected during anaerobic digestion allowed implementation of a two-phase mathematical model.The thirteen parameters of the model were estimated with particle swarm optimisation method in Matlab.The model was able to simulate the biomethane production kinetics and variation of volatile fatty acid concentration.

Multiobjective optimal dispatch of microgrid based on analytic hierarchy process and quantum particle swarm optimization

Owing to the rapid development of microgrids(MGs)and growing applications of renewable energy resources,multiobjective optimal dispatch of MGs need to be studied in detail.In this study,a multiobjective optimal dispatch model is developed for a standalone MG composed of wind turbines,photovoltaics,diesel engine unit,load,and battery energy storage system.The economic cost,environmental concerns,and power supply consistency are expressed via subobjectives with varying priorities.Then,the analytic hierarchy process algorithm is employed to reasonably specify the weight coefficients of the subobjectives.The quantum particle swarm optimization algorithm is thereafter employed as a solution to achieve optimal dispatch of the MG.Finally,the validity of the proposed model and solution methodology are con firmed by case studies.This study provides refere nee for mathematical model of multiojective optimizati on of MG and can be widely used in current research field.

基于QPSO-LightGBM网络资产脆弱性评估模型

为有效减少网络安全事件造成的损失,并对高风险网络资产进行漏洞评估,提出了一种基于量子粒子群轻量级梯度升降算法(QPSO-LightGBM:Quantum Particle Swarm Optimization-Light Gradient Boosting Machine)的多分类预测模型。该模型通过对少量过采样技术(MOTE:Minority Oversampling)进行合成从而达到数据平衡,采用量子粒子群算法(QPSO:Quantum Particle Swarm Optimization)实现参数的自动最优化,并使用LightGBM进行建模,进而实现网络资产的多分类预测。为验证模型的有效性,将所提模型与其他算法模型进行了比对,实验结果表明,该模型在各类预测性能指标上都取得了较好的效果。

Mooring System Optimisation and Effect of Different Line Design Variables on Motions of Truss Spar Platforms in Intact and Damaged Conditions

This paper presents the effect of mooring diameters, fairlead slopes and pretensions on the dynamic responses of a truss spar platform in intact and damaged line conditions. The platform is modelled as a rigid body with three degrees-of-freedom and its motions are analysed in time-domain using the implicit Newmark Beta technique. The mooring restoring force-excursion relationship is evaluated using quasi-static approach. MATLAB codes DATSpar and QSAML, are developed to compute the dynamic responses of truss spar platform and to determine the mooring system stiffness. To eliminate the conventional trial and error approach in the mooring system design, a numerical tool is also developed and described in this paper for optimising the mooring configuration. It has a graphical user interface and includes regrouping particle swarm optimisation technique combined with DATSpar and QSAML. A case study of truss spar platform with ten mooring lines is analysed using this numerical tool. The results show that optimum mooring system design benefits the oil and gas industry to economise the project cost in terms of material, weight, structural load onto the platform as well as manpower requirements. This tool is useful especially for the preliminary design of truss spar platforms and its mooring system.

Remarks on the Efficiency of Bionic Optimisation Strategies

Bionic optimisation is one of the most popular and efficient applications of bionic engineering. As there are many different approaches and terms being used, we try to come up with a structuring of the strategies and compare the efficiency of the different methods. The methods mostly proposed in literature may be classified into evolutionary, particle swarm and artificial neural net optimisation. Some related classes have to be mentioned as the non-sexual fern optimisation and the response surfaces, which are close to the neuron nets. To come up with a measure of the efficiency that allows to take into account some of the published results the technical optimisation problems were derived from the ones given in literature. They deal with elastic studies of frame structures, as the computing time for each individual is very short. General proposals, which approach to use may not be given. It seems to be a good idea to learn about the applicability of the different methods at different problem classes and then do the optimisation according to these experiences. Furthermore in many cases there is some evidence that switching from one method to another improves the performance. Finally the identification of the exact position of the optimum by gradient methods is often more efficient than long random walks around local maxima.

基于QPSO优化ELM算法的井下快速定位方法研究

近年来,基于WiFi的井下人员定位算法受到了广泛关注,为进一步提高定位算法的速度和精度,提出了一种基于量子粒子群(QPSO)优化极限学习机(ELM)算法的井下人员快速定位方法。首先将优化的K均值聚类算法(K-means)引入定位流程,通过对位置指纹库进行聚类划分,降低单次识别需要的时间;其次利用QPSO算法优越的优化能力,提高ELM算法的定位精度,最终实现对井下人员的高效、准确定位。仿真实验结果表明,提出的基于量子粒子群优化极限学习机的井下快速定位方法,能够提高定位的精度和效率,具有较高的实用价值和工程意义。

多场景下基于AHP-EWM的人体健康状态评估模型研究

为解决人体健康评估方法个性化监测不足的问题以及在满足不同场景下健康状态精细化评估的需求,需要一种基于多场景的人体健康状态评估方法来实现长期自动化监测。提出一种基于层次分析法(AHP)和熵权法(EWM)组合的多场景人体健康状态评估模型。首先采集人体在运动、休息、工作/学习和娱乐等4种不同场景下的健康监测指标数据,构建相应的评估指标体系。然后分别根据评估指标计算出AHP和EWM权重,再采用量子粒子群优化(QPSO)算法对AHP和EWM中的主客观权重进行分配,以确保评价指标占比的客观性。最后通过模糊综合评价法对人体健康状态进行评估和量化,并利用实际监测数据对方法的可靠性和稳定性进行验证。实验结果表明,在4种场景下所提方法的综合得分分别为63.78、59.83、58.71和59.21,表明在不同场景下该模型都具有较好的准确性和稳定性。根据评估结果,对测试者的身体状态评价结果进行分析,并给出一些健康建议。所提模型可全面了解人体在不同场景下的健康状况,并为人们提供科学的健康指导,从而为健康管理和疾病预防提供科学依据。

基于多策略混合改进MVO算法的光伏多峰MPPT研究

光伏阵列的电功率输出在局部遮荫工况下具有多峰特性,且随外界环境的变化而变化.为实现高效电能输出,利用多元宇宙优化(MVO)算法在解决低维度、小规模优化问题中的突出优势进行最大功率点跟踪(MPPT),并融合多种策略对其存在的缺陷进行改进.采用拉丁超立方抽样策略初始化宇宙种群,并对按照轮盘赌策略随机交换的宇宙实施柯西变异,提高宇宙种群的多样性.同时引入莱维飞行式量子粒子群优化(QPSO)算法,且对虫洞存在概率及旅行距离率进行自适应调整,以增强算法的全局勘探及局部开发能力.Matlab仿真结果表明,相比其他算法,采用该算法的MPPT时间减少了45%以上,精度亦有所提高,从而具有更优的MPPT性能,可有效提高光伏发电效率.

基于改进PSO-BP神经网络的热采管柱应力预测

稠油热采过程中,油套管柱由于在温度、地层等多重载荷作用下发生塑性形变进而导致断裂或失效。文中根据热采管柱高温服役工况,引入异步变化学习因子和自适应权重建立输入参数为注汽温度、井深、非均匀系数和水泥环温度,输出参数为套管应力的改进PSO-BP模型。文中以N80热采套管为例,选取260、280、300、320、340℃5种温度工况下有限元模拟结果作为训练数据,对比BP模型、GA-BP模型、MEA-BP模型、PSO-BP模型和改进PSO-BP模型在300℃工况温度下井深200、300、400、500、600、700 m处套管应力的预测值和试验值、有限元计算值。结果表明:改进PSO-BP模型预测的应力与试验值最接近,最大和最小误差分别为2.69%和0.06%。最后从训练数据、预测误差、计算时间等方面对建立的改进PSO-BP模型进行了评价,为热采管柱服役过程中的强度安全分析提供智能高效的模型。

一种基于QPSO-RVM的模拟电路故障预测方法

提出了一种可应用于模拟电路故障预测的方法。通过提取被测电路的频域响应信号,计算皮尔逊相关系数,从而表征电路元件的健康度;在获取元件在不同时间点的健康度数据的基础上,推导出电路元件发生故障时的健康度阈值;将经量子粒子群算法优化的相关向量机算法用于故障预测,预测各个时间点的元件健康度变化轨迹并估计模拟电路的剩余有用寿命。该预测方法计算简单、通用性强,适用于实时预测。故障预测仿真实验与实例实验证明了方法的有效性与先进性。

基于QPSO-RBF的瓦斯涌出量预测模型

为了提高径向基(RBF)网络预测瓦斯涌出量的泛化能力,提出QPSO-RBF模型。该模型采用量子粒子群(QPSO)算法优化RBF网络隐层基函数中心、扩展系数以及输出权等初始参数,将网络参数编码为QPSO学习算法中的粒子个体,在全局空间中搜索最优适应值参数。其中,RBF网络选取5-3-1的精简结构,采用5个变量作为影响因子预测瓦斯涌出量。结果表明,经QPSO优化后的RBF网络模型预测结果稳定且唯一,其泛化指标平均相对变动值(ARV)为0.012 2。与PSO-RBF、RBF模型预测结果比较,QPSO-RBF模型的泛化能力和网络训练速度优于前2种;预测精度约为PSO-RBF模型的1.5倍、RBF模型的4倍。

基于QPSO的数据聚类

在K-Means聚类、PSO聚类、K-Means和PSO混合聚类(KPSO)的基础上,研究了基于量子行为的微粒群优化算法(QPSO)的数据聚类方法,并提出利用K-Means聚类的结果重新初始化粒子群,结合QPSO的聚类算法,即KQPSO。介绍了如何利用上述算法找到用户指定的聚类个数的聚类中心。聚类过程都是根据数据之间的Euclidean(欧几里得)距离。K-Means算法、PSO算法和QPSO算法的不同在于聚类中心向量的“进化”上。最后使用三个数据集比较了上面提到的五种聚类方法的性能,结果显示基于QPSO算法的数据聚类性能比一般PSO算法更好。

QPSO算法优化BP网络的网络流量预测

网络流量预测对于大规模网络的规划设计和网络资源管理等方面都具有积极的意义,是网络流量工程重要组成部分。结合QPSO算法和BP神经网络的优势,采用QPSO算法对BP神经网络的权值和阈值进行优化,并利用历史记录训练BP网络。仿真实验表明,与PSO训练的BP网络以及直接用BP网络进行预测的模型相比,基于QPSO训练的BP网络流量预测模型具有更好的预测能力。

基于QPSO算法的RBF神经网络参数优化仿真研究

针对粒子群优化(PSO)算法搜索空间有限,容易陷入局部最优点的缺陷,提出一种以量子粒子群优化(QPSO)算法为基础的RBF神经网络训练算法,将RBF神经网络的参数组成一个多维向量,作为算法中的粒子进行进化,由此在可行解空间范围内搜索最优解。实例仿真表明,该学习算法相比于传统的学习算法计算简单,收敛速度快,并由于其算法模型的自身特性比基于PSO的学习算法具有更好的全局收敛性能。

基于DGA的QPSO-BP模型变压器故障诊断方法研究

为了提高变压器故障诊断的准确率,提出一种基于量子粒子群优化BP神经网络(quantum particle swarm optimized BP neural network,QPSO-BP)的故障诊断模型。在该算法中,用量子位的概率幅表示种群中各粒子的当前位置,用量子旋转门实现粒子位置的移动,用量子非门进行变异操作,以获取BP神经网络的权、阈值优化参数,最终实现了变压器故障诊断模型的构建。对故障DGA样本的诊断实例表明,与粒子群优化BP网络(particle swarm optimized BP neural network,PSO-BP)法、BPNN法以及IEC三比值法相比,QPSO-BP算法具有更高的诊断正确率,从而实现了变压器故障模式的有效识别。

云计算环境下基于ABC-QPSO算法的资源调度模型

为了提高云计算资源的利用率,保证节点负载均衡,提出一种人工蜂群算法和量子粒子群算法相融合的云计算资源调度模型(ABC-QPSO)。首先将人工蜂群算法的搜索算子作为变异算子引入到量子粒子群算法中,以解决量子粒子群算法早熟收敛缺陷,然后以任务完成时间最短作为量子粒子群的适应度函数对云计算资源调度进行优化,最后在Cloud Sim平台上对ABC-QPSO的性能进行测试。结果表明,ABC-QPSO算法不仅克服了QPSO算法的不足,而且有效缩短了任务的完成时间,提高了云计算资源利用率,适合于进行大规模任务的云计算资源调度。

运用QPSO算法进行系统辨识的研究

引入了一种广泛而实用的方法——基于量子行为的粒子群算法的理论应用于系统辨识领域,QPSO算法不仅参数个数少,随机性强,并且能覆盖所有解空间,保证算法的全局收敛性。仿真实验结果表明,QPSO算法具有比GA算法及PSO算法更强的线性系统辨识能力和非线性系统辨识能力。

基于改进的QPSO训练BP网络的网络流量预测

为了提高网络流量预测的精度,采用一种改进的QPSO算法训练BP神经网络对网络流量数据的时间序列进行建模预测。针对标准的QPSO算法不可避免地出现早熟的不足,提出一种新的基于参数自适应的QPSO算法,较好地避免了粒子群的早熟,提高了算法的全局收敛性能。仿真实验结果表明,与PSO训练的BP网络、QPSO训练的BP网络作为预测模型相比,该模型具有更高的预测精度及很好的稳定性。