Yarn Quality Prediction for Small Samples Based on AdaBoost Algorithm

In order to solve the problems of weak prediction stability and generalization ability of a neural network algorithm model in the yarn quality prediction research for small samples,a prediction model based on an AdaBoost algorithm(AdaBoost model) was established.A prediction model based on a linear regression algorithm(LR model) and a prediction model based on a multi-layer perceptron neural network algorithm(MLP model) were established for comparison.The prediction experiments of the yarn evenness and the yarn strength were implemented.Determination coefficients and prediction errors were used to evaluate the prediction accuracy of these models,and the K-fold cross validation was used to evaluate the generalization ability of these models.In the prediction experiments,the determination coefficient of the yarn evenness prediction result of the AdaBoost model is 76% and 87% higher than that of the LR model and the MLP model,respectively.The determination coefficient of the yarn strength prediction result of the AdaBoost model is slightly higher than that of the other two models.Considering that the yarn evenness dataset has a weaker linear relationship with the cotton dataset than that of the yarn strength dataset in this paper,the AdaBoost model has the best adaptability for the nonlinear dataset among the three models.In addition,the AdaBoost model shows generally better results in the cross-validation experiments and the series of prediction experiments at eight different training set sample sizes.It is proved that the AdaBoost model not only has good prediction accuracy but also has good prediction stability and generalization ability for small samples.

Quantitative algorithm for airborne gamma spectrum of large sample based on improved shuffled frog leaping-particle swarm optimization convolutional neural network

In airborne gamma ray spectrum processing,different analysis methods,technical requirements,analysis models,and calculation methods need to be established.To meet the engineering practice requirements of airborne gamma-ray measurements and improve computational efficiency,an improved shuffled frog leaping algorithm-particle swarm optimization convolutional neural network(SFLA-PSO CNN)for large-sample quantitative analysis of airborne gamma-ray spectra is proposed herein.This method was used to train the weight of the neural network,optimize the structure of the network,delete redundant connections,and enable the neural network to acquire the capability of quantitative spectrum processing.In full-spectrum data processing,this method can perform the functions of energy spectrum peak searching and peak area calculations.After network training,the mean SNR and RMSE of the spectral lines were 31.27 and 2.75,respectively,satisfying the demand for noise reduction.To test the processing ability of the algorithm in large samples of airborne gamma spectra,this study considered the measured data from the Saihangaobi survey area as an example to conduct data spectral analysis.The results show that calculation of the single-peak area takes only 0.13~0.15 ms,and the average relative errors of the peak area in the U,Th,and K spectra are 3.11,9.50,and 6.18%,indicating the high processing efficiency and accuracy of this algorithm.The performance of the model can be further improved by optimizing related parameters,but it can already meet the requirements of practical engineering measurement.This study provides a new idea for the full-spectrum processing of airborne gamma rays.

Scaling up the DBSCAN Algorithm for Clustering Large Spatial Databases Based on Sampling Technique

Clustering, in data mining, is a useful technique for discovering interesting data distributions and patterns in the underlying data, and has many application fields, such as statistical data analysis, pattern recognition, image processing, and etc. We combine sampling technique with DBSCAN algorithm to cluster large spatial databases, and two sampling based DBSCAN (SDBSCAN) algorithms are developed. One algorithm introduces sampling technique inside DBSCAN, and the other uses sampling procedure outside DBSCAN. Experimental results demonstrate that our algorithms are effective and efficient in clustering large scale spatial databases.

Iterative Learning Fault Diagnosis Algorithm for Non-uniform Sampling Hybrid System

For a class of non-uniform output sampling hybrid system with actuator faults and bounded disturbances,an iterative learning fault diagnosis algorithm is proposed.Firstly,in order to measure the impact of fault on system between every consecutive output sampling instants,the actual fault function is transformed to obtain an equivalent fault model by using the integral mean value theorem,then the non-uniform sampling hybrid system is converted to continuous systems with timevarying delay based on the output delay method.Afterwards,an observer-based fault diagnosis filter with virtual fault is designed to estimate the equivalent fault,and the iterative learning regulation algorithm is chosen to update the virtual fault repeatedly to make it approximate the actual equivalent fault after some iterative learning trials,so the algorithm can detect and estimate the system faults adaptively.Simulation results of an electro-mechanical control system model with different types of faults illustrate the feasibility and effectiveness of this algorithm.

Optimization of Process Parameters for Cracking Prevention of UHSS in Hot Stamping Based on Hammersley Sequence Sampling and Back Propagation Neural Network-Genetic Algorithm Mixed Methods

In order to prevent cracking appeared in the work-piece during the hot stamping operation,this paper proposes a hybrid optimization method based on Hammersley sequence sampling( HSS),finite analysis,backpropagation( BP) neural network and genetic algorithm( GA). The mechanical properties of high strength boron steel are characterized on the basis of uniaxial tensile test at elevated temperatures. The samples of process parameters are chosen via the HSS that encourages the exploration throughout the design space and hence achieves better discovery of possible global optimum in the solution space. Meanwhile, numerical simulation is carried out to predict the forming quality for the optimized design. A BP neural network model is developed to obtain the mathematical relationship between optimization goal and design variables,and genetic algorithm is used to optimize the process parameters. Finally,the results of numerical simulation are compared with those of production experiment to demonstrate that the optimization strategy proposed in the paper is feasible.

Potential-Decomposition Strategy in Markov Chain Monte Carlo Sampling Algorithms

We introduce the potential-decomposition strategy （PDS）, which can be used in Markov chain Monte Carlo sampling algorithms. PDS can be designed to make particles move in a modified potential that favors diffusion in phase space, then, by rejecting some trial samples, the target distributions can be sampled in an unbiased manner. Furthermore, if the accepted trial samples are insumcient, they can be recycled as initial states to form more unbiased samples. This strategy can greatly improve efficiency when the original potential has multiple metastable states separated by large barriers. We apply PDS to the 2d Ising model and a double-well potential model with a large barrier, demonstrating in these two representative examples that convergence is accelerated by orders of magnitude.

Algorithm-based arterial blood sampling recognition increasing safety in point-of-care diagnostics

AIM To detect blood withdrawal for patients with arterial blood pressure monitoring to increase patient safety and provide better sample dating.METHODS Blood pressure information obtained from a patient monitor was fed as a real-time data stream to an experimental medical framework. This framework was connected to an analytical application which observes changes in systolic, diastolic and mean pressure to determine anomalies in the continuous data stream. Detection was based on an increased mean blood pressure caused by the closing of the withdrawal three-way tap and an absence of systolic and diastolic measurements during this manipulation. For evaluation of the proposed algorithm, measured data from animal studies in healthy pigs were used.RESULTS Using this novel approach for processing real-time measurement data of arterial pressure monitoring, the exact time of blood withdrawal could be successfully detected retrospectively and in real-time. The algorithm was able to detect 422 of 434(97%) blood withdrawals for blood gas analysis in the retrospective analysis of 7 study trials. Additionally, 64 sampling events for other procedures like laboratory and activated clotting time analyses were detected. The proposed algorithm achieved a sensitivity of 0.97, a precision of 0.96 and an F1 score of 0.97.CONCLUSION Arterial blood pressure monitoring data can be used toperform an accurate identification of individual blood samplings in order to reduce sample mix-ups and thereby increase patient safety.

A New Random Sampling Method and Its Application in Improving Progressive BKZ Algorithm

Random sampling algorithm was proposed firstly by Schnorr in 2003 to find short lattice vectors,as an alternative to enumeration.The follow-up developments in random sampling were mainly proposed by Fukase and Kashiwabara in 2015 and Aono and Nguyen in 2017.Although they extended the sampling space compared to Schnorr's work through the natural number representation,they did not show how to sample specifically in practice and what vectors should be sampled,in order to find short enough lattice vectors.In this paper,the authors firstly introduce a practical random sampling algorithm under some reasonable assumptions which can find short enough lattice vectors efficiently.Then,as an application of this new random sampling algorithm,the authors show that it can improve the performance of progressive BKZ algorithm in practice.Finally,the authors solve the Darmstadt's Lattice Challenge and get a series of new records in the dimension from 500 to 825,using the improved progressive BKZ algorithm.

The study and application of PTR algorithm on recognizing various structure samples

In this paper,four pattern recognition methods are set forth.Based on plane projection of samples and analysis of typical samples along with the few pattern recognition methods,the PTR algorithm for recognizing various structure samples is proposed.Also two examples are given and these show the PTR algorithm is effective.

Data-driven evolutionary sampling optimization for expensive problems

Surrogate models have shown to be effective in assisting evolutionary algorithms(EAs)for solving computationally expensive complex optimization problems.However,the effectiveness of the existing surrogate-assisted evolutionary algorithms still needs to be improved.A data-driven evolutionary sampling optimization(DESO)framework is proposed,where at each generation it randomly employs one of two evolutionary sampling strategies,surrogate screening and surrogate local search based on historical data,to effectively balance global and local search.In DESO,the radial basis function(RBF)is used as the surrogate model in the sampling strategy,and different degrees of the evolutionary process are used to sample candidate points.The sampled points by sampling strategies are evaluated,and then added into the database for the updating surrogate model and population in the next sampling.To get the insight of DESO,extensive experiments and analysis of DESO have been performed.The proposed algorithm presents superior computational efficiency and robustness compared with five state-of-the-art algorithms on benchmark problems from 20 to 200 dimensions.Besides,DESO is applied to an airfoil design problem to show its effectiveness.

Optimization of Well Position and Sampling Frequency for Groundwater Monitoring and Inverse Identification of Contamination Source Conditions Using Bayes’Theorem

Coupling Bayes’Theorem with a two-dimensional(2D)groundwater solute advection-diffusion transport equation allows an inverse model to be established to identify a set of contamination source parameters including source intensity(M),release location(0 X,0 Y)and release time(0 T),based on monitoring well data.To address the issues of insufficient monitoring wells or weak correlation between monitoring data and model parameters,a monitoring well design optimization approach was developed based on the Bayesian formula and information entropy.To demonstrate how the model works,an exemplar problem with an instantaneous release of a contaminant in a confined groundwater aquifer was employed.The information entropy of the model parameters posterior distribution was used as a criterion to evaluate the monitoring data quantity index.The optimal monitoring well position and monitoring frequency were solved by the two-step Monte Carlo method and differential evolution algorithm given a known well monitoring locations and monitoring events.Based on the optimized monitoring well position and sampling frequency,the contamination source was identified by an improved Metropolis algorithm using the Latin hypercube sampling approach.The case study results show that the following parameters were obtained:1)the optimal monitoring well position(D)is at(445,200);and 2)the optimal monitoring frequency(Δt)is 7,providing that the monitoring events is set as 5 times.Employing the optimized monitoring well position and frequency,the mean errors of inverse modeling results in source parameters(M,X0,Y0,T0)were 9.20%,0.25%,0.0061%,and 0.33%,respectively.The optimized monitoring well position and sampling frequency canIt was also learnt that the improved Metropolis-Hastings algorithm(a Markov chain Monte Carlo method)can make the inverse modeling result independent of the initial sampling points and achieves an overall optimization,which significantly improved the accuracy and numerical stability of the inverse modeling results.

MULTIPLE FREQUENCIES ESTIMATION OF SIGNAL WITH SUB-SAMPLING

Based on time delay technology and MUSIC algorithm, a novel estimating multiple frequencies approach of signal with sampling rate which is least Nyquist sampling rate is presented in this paper. With choosing delay time properly, the estimated frequencies are unambiguous. Computer simulation confirms its availability.

改进RRT算法的采摘机械臂路径规划

针对采用传统的快速随机扩展树(RRT)算法的采摘机械臂在果园工作环境中搜索路径时间长,最终路径不平滑、拐点多等问题,提出了一种改进的RRT避障算法。改进的算法采用高斯采样策略,减少了采样的随机性,避免产生更多不必要的随机树,增加规划的导向性;再添加A*代价函数去除路径的冗余点,最后使用贪婪算法简化路径,减少拐点,让机械臂可以快速、准确、平稳地沿着最佳路径运动到目标点。仿真表明,改进后的算法有效地减少了路径规划的时间,缩短了路径长度,具有良好的可行性和有效性。

基于迁移学习的轨道交通特殊OD客流预测研究

客流预测一直是轨道交通运营公司关注的重点,由于受到运输能力的限制等因素影响,部分OD的实际客流数据与真实需求有偏差,出现异常或者样本缺失,从而造成总体样本量偏小,直接采用这些样本进行预测会明显影响预测精度,但通过还原样本值增加样本量难度太大。根据上述特点选择基于实例的迁移学习,先确定源域的对象和范围,从源域中选择合适的样本补充到总体样本中,共同组成最终的训练样本数据集,完成迁移学习。同时选择改进的Boost算法,通过误差调整样本权重,不断迭代,得到最终的预测模型。结果表明:基于实例的迁移学习结合改进Boost算法的预测精度要好于传统集成学习、ARIMA模型、多元回归模型,为轨道交通运营公司对特定OD的客流预测提供新的有益尝试。

结合注意力和多路径融合的实时肺结节检测算法

现有单阶段目标检测算法在肺结节检测中结节检出不敏感,卷积神经网络(CNN)在特征提取时多次上采样导致微小结节特征提取困难、检测效果差,并且现存肺结节检测算法模型复杂,不利于实际应用部署落地。针对上述问题,提出一种结合注意力机制和多路径融合的实时肺结节检测算法,并在此基础上改进上采样算法,提升肺部结节的检测精度和模型推理速度,且模型的权重小容易部署。首先,在特征提取的主干网络部分融合通道和空间的混合注意力机制;其次,改进采样算法,提高生成特征图的质量;最后在加强特征提取网络部分,在不同路径之间建立通道,实现深层和浅层特征的融合,将不同尺度的语义和位置信息融合。在LUNA16数据集的实验结果表明,相较于原始YOLOv5s算法,所提算法的精确率、敏感度和平均精度分别提升9.5、6.9和8.7个百分点,帧率达到131.6 frame/s,模型权重文件仅有14.2 MB,表明了所提算法可以实时检测肺结节,并且精度远高于YOLOv3和YOLOv8等现有单阶段检测算法。

编制价格指数的爬虫数据抽样方法研究

文章针对全量爬虫数据编制价格指数成本高的问题,提出了一种抽样方法。该方法采用“大数据—小数据”思想,在基期通过网络爬虫技术全量抓取电商平台的商品交易数据,形成抽样框;在连续性调查中采用抽样技术,根据分层抽样思想,运用聚类算法及其轮廓系数实现总体数据分层,并通过不等概率随机抽样获取各层代表性样本;考虑到连续性调查中入选样本存在无回答现象,提出正式和备选样本思路,针对每个正式样本,采用最近邻匹配法挑选若干个备选样本,当正式样本无回答时,以备选样本作为替补来完成价格指数编制。以天猫商城粮油品类为例进行验证,结果表明:在抓取的数据中,基期全量爬虫数据有18351条,第2—8期连续性调查的平均抽样比为10.18%,抽样的平均相对误差为0.59%,说明该方法是可行的。

基于级联碰撞缺陷数据库的源项对辐照微结构演化影响团簇动力学模拟研究

团簇动力学(CD)方法是模拟核材料在高能粒子辐照下微观结构演化的重要方法之一,源项是团簇动力学方法的关键输入。经典CD方法中源项通常采用经验拟合得到,未能充分利用原子尺度获得的初始缺陷信息。随着分子动力学等方法的发展,级联碰撞缺陷数据库大为丰富,结合初级离位原子(PKA)能谱足以得到更为合理的源项。由于级联碰撞缺陷数据库的能量值数量相对于准连续PKA能谱仍然偏少,本文提出了5种从准连续PKA能谱得到级联能量分立值的抽样算法,并基于团簇动力学方法模拟低剂量中子辐照纯钨实验对算法进行了验证和比较。

基于不平衡数据的物联网异常流量检测

为应对数据类别不平衡问题,导致物联网异常流量检测模型性能低下,提出一种基于不平衡数据的物联网异常流量检测方法.首先,采用基于马氏距离(Mahalanobis distance,MD)的K-means SMOTE-ENN算法生成无噪声的数据,以有效实现数据样本分布均衡.其次,针对异常流量检测模型性能低下,构建了卷积神经网络(convolutional neural network,CNN)和双向长短期记忆网络(bi-directional long short-term memory,BiLSTM)相结合的模型,提取异常流量的局部卷积特征以及关键特征.最后,通过全连接层和分类器进行分类.实验结果显示,相较于现有异常流量检测方法,所提出的方法在准确率、召回率、精确率和F1值等评价指标上均取得显著提升.该模型能够准确识别流量中的异常行为,准确率高达99.43%.

一种发射系下旋转弹的捷联惯导算法

为了提高旋转弹的姿态测量精度,提出了一种发射系下的旋转弹捷联导航算法。首先,推导了发射系导航更新方程;然后,利用传感器测量的前周期角速率对姿态更新算法进行优化,推导了基于角速率的旋转矢量更新算法及其算法误差;最后,采用陀螺仪标度因数的提升方法,实现了捷联惯导姿态测量精度有效的提高。通过对双子样、优化双子样算法在自旋运动条件下的数学仿真、基于转台的半实物仿真试验进行了验证。试验结果表明:在自旋机动条件下,相较于传统双子样算法,基于发射系的旋转弹捷联导航算法具有更高的精度,验证了算法在工程应用实现的正确性和有效性。

基于改进蚁群算法的对抗样本生成模型

传统的文本生成对抗方法主要采用位置置换、字符替换等方式,耗费时间较长且效果较差。针对以上问题,该文提出一种基于改进蚁群算法的对抗样本生成模型IGAS(Improved ant colony algorithm to Generate Adversarial Sample),利用蚁群算法的特点生成对抗样本,并利用类形字进行优化。首先,构建城市节点群,利用样本中的词构建城市节点群;然后对原始输入样本,利用改进的蚁群算法生成对抗样本;再针对生成结果,通过构建的中日类形字典进行字符替换,生成最终的对抗样本;最后在黑盒模式下进行对抗样本攻击实验。实验在情感分类、对话摘要生成、因果关系抽取等多种领域验证了该方法的有效性。