登录

双语推荐：自学习阈值

进入期刊文章
针对当前机械在线监测系统报警难以实现机械故障早期预警问题,提出一种智能预警方法。基于在线监测系统大量监测数据统计分析,采用动态的自学习阈值算法计算预警阈值,并应用l1趋势滤波技术消除随机误差获取滤波后的趋势。应用动态自学习阈值替代监测系统中的常规报警阈值,比较自学习预警阈值与滤波后的趋势,实现了机械故障早期预警。工程实例表明,该方法能够对机械故障实现早期预警,对预防机械事故的发生有重要的作用。
Because the alarm mode of current online monitoring systems is hard to realize early-warning a mechanical fault,an early warning methodology was proposed here.Based on statistical analysis of mass data in online monitoring systems,advantages of the dynamic self learning threshold algorithm were taken to calculate a warning threshold,and the l1 trend filtering technology was used to gain the filtered trend eliminating random errors.With dynamic self-learning threshold instead of general alarm threshold in monitoring systems,early warning of a fault could be acquired by comparing the self-learning warning threshold and the filtered trend.It was shown from engineering examples that early warning of a mechanical fault can be achieved with the proposed method,it plays an important role in preventing occurrence of mechanical fault.

[ 可能符合您检索需要的词汇 ]

进入期刊文章
针对混合高斯背景减除法在运动目标检测应用中存在的不足,进行了以下两个方面的改进:第一,通过在混合高斯模型匹配中引入自适应匹配阈值的方法,解决由噪声或光照引起的误判问题;第二,在模型学习方面,采用不同的权重学习速率以检测静态背景区域,并提高模型的自适应性。实验结果表明,与传统的混合高斯模型的运动目标检测方法相比,改进后的方法在背景误判、场景适应性方面都有所改善。
This paper makes improvements on moving target detection method which is based on mixture Gaussian mod-el, specifically in two areas:the use of adaptive matching threshold solves the problem of misdetection of moving targets caused by noise or illumination change;and in terms of model learning, it uses different learning rate to detect static back-ground areas, improving the adaptiveness of model. Compared to the moving targets detection approach based on conven-tional mixture Gaussian model, the improved methods have significantly solved the problems of misdetection of moving object, and the adaptiveness of model.

[ 可能符合您检索需要的词汇 ]

进入期刊文章
近年来,Twitter搜索在社交网络领域引起越来越多学者的关注.尽管排序学习可以融合Twitter中丰富的特征,但是训练数据的匮乏,会降低排序学习的性能.直推式学习作为一种常用的半监督学习方法,在解决训练数据的稀少性中发挥着重要的作用.由于在直推式学习的迭代过程中会生成噪音,基于聚类的直推式学习方法被提出.在基于聚类的直推式学习方法中有两个重要的参数,分别为聚类的阈值以及聚类文档的数量.在原有工作的基础上,提出使用另外一种不同的聚类算法.大量在标准TREC数据集Tweets11上的实验表明,聚类的阈值以及聚类过程中文档数量的选择都会对模型的检索性能产生影响.另外,也分析了基于聚类的直推式学习模型的鲁棒性在不同查询集上的表现.最后,引入名为簇凝聚度的质量控制因子,提出了一种基于聚类的自适应的直推式方法来实现Twitter检索.实验结果表明,基于聚类的自适应学习算法具有更好的鲁棒性.
Recently, Twitter search has drawn much attention of researchers in social networks. Although rich features of Twitter can be incorporated into rank learning, the retrieval effectiveness can be hurt by the lack of training data. Transductive learning, as a common semi-supervised learning method, has been playing an import role in dealing with the lacking of training data. Due to the fact that noise is generated during the iterative process of transductive learning, a clustering-based transductive method is proposed. There exist two important parameters in the clustering-based transductive approach, namely the threshold of clustering and the number of the documents that will be clustered. This paper extends the method by utilizing a different clustering algorithm. As shown by extensive experiments on the standard TREC Tweets11 collection, both of the two parameters have an effect on the retrieval effectiveness. Furthermore, the robustness of the clustering-based transduction a

[ 可能符合您检索需要的词汇 ]

进入期刊文章
命名实体识别是信息抽取中的一项基础性任务,如何利用丰富的未标注语料来提高实体识别的指标是该领域一个重要的研究方向。基于条件随机场提出一种将主动学习与自学习相结合的方法——SACRF,通过设置置信度函数和2-Gram频度阈值来选取样本,并采用人工与自动相结合的方式进行标注来扩展训练语料。实验表明,该方法在提高实体识别的精确率和召回率的同时,能够显著地降低人工标注的工作量。
Named Entity Recognition (NER)is a basic task in information extraction,and it is an important research direction in this domain to use the abundant unlabeled corpus to improve the performance of NER system.An approach combining self-training with active learning based on CRF (SACRF)is proposed.It selected samples by setting the threshold of confidence and 2-Gram frequency,and expanded the training set by annotating the unlabeled corpus manually and automatically.The experiments revealed that this approach can not only improve the precision and recall of NER system,but also reduce the manually annotation efforts greatly.

[ 可能符合您检索需要的词汇 ]

进入期刊文章
该文将神经网络应用于抽油系统的故障诊断,根据泵功图的几何特征提取特征值作为BP神经网络的输入信号,利用自适应性以及线性映射能力,建立抽油系统输入的故障信息与输出的故障类型间的映射。通过对大量故障样本的学习将知识以权值和阈值的形式存储于网格中,最终输出抽油系统的故障类型。通过实例分析,模型具有比较高的准确性和可行性。
Neural network is used in the fault diagnosis of suck rod pumping system .The characteristic value is obtained based the geometric characteristic of the pump dynamometer card as input signals of back propagation( BP) neural networks .The relations between the network fault information and fault patterns are established utilizing the self-adaptation and nonlinearity mapping functions of the neural network . The knowledge in nets is kept in the form of weigh and threshold by learning from the fault samples , and the outputs of nets are typical models of the fault .After numerical analysis , the results indicate the feasibility and veracity of this method .

[ 可能符合您检索需要的词汇 ]

进入期刊文章
催化裂化反应-再生系统是一个高度非线性和强耦合的操作系统,用传统建模方法很难描述。鉴于人工神经网络（ANN）非线性预测和自学习自适应能力强,而遗传算法（GA）全局寻优能力强的特点,将两者结合,先通过GA寻得BP神经网络最优的权值和阈值初值,再赋予BP,从而改善BP模型随机不确定选择初值的方法,提高其映射精度。以某炼油厂2.8Mt/a MIP装置反应-再生系统为研究对象,选取第一反应区温度、第二反应区温度、第一再生器温度、第二再生器温度、反应器压力、再生器压力等6个变量为神经网络的输入变量,汽油产率为输出变量,建立6-11-1的BP神经网络,并采用GA来对BP神经网络的权值和阈值进行优化。结果表明,未经GA优化时BP神经网络对催化裂化汽油产率的预测数据的均方误差为5.16,而经GA优化后预测数据的均方误差为4.92。
The system of reaction and generation unit of RFCCU is a highly non-linear and strong coupled operation system and is too hard to be described by traditional model. The combination of the artificial neural network (ANN)with strong nonlinear prediction and self-learning ability and the genet-ic algorithm (GA)with global optimization ability provides a promising way to solve the problem. The optimal initial weights and threshold value are calculated by GA for the BP neural network firstly and feeded back to BP model to improve the method for random uncertain choice of initial value and the map-ping accuracy. In a practical application of this method for a 2.8 Mt/a MIP unit,a 6-11-1 type of BP neural network where the GA is used to optimize the weights and values of the BP network was estab-lished using the temperatures of two reaction zones and two regenerators along with the pressures of the reactor and regenerator as six input variables to predict the output variable gasoline yield. Th

[ 可能符合您检索需要的词汇 ]

进入期刊文章
针对传统车辆检测方法定位精度不高的问题,提出一种基于多特征融合的前向车辆检测方法。采用基于直方图分析和自适应双阈值的方法分别实现阴影和边缘特征的准确分割,并通过阴影和边缘特征的综合分析,生成车辆假设区域。利用对称性、纹理和轮廓匹配度3个特征融合得到的综合特征对获得的车辆假设区域进行验证,剔除其中的误检区域。实验结果证明,该方法能在不同光照条件下自适应地进行车辆检测,检测率可达92%以上,且在检测率和误检率2项指标上均优于传统基于学习的方法。
A forward vehicle detection method based on multi-feature fusion is proposed in order to improve the accuracy of vehicle detection. The shadow and edge features of vehicle are segmented accurately by using histogram analysis method and adaptive dual-threshold method respectively. The initial candidates are generated by combining edge and shadow features and these initial candidates are further verified by using an integrated feature based on the fusion of symmetry, texture and shape matching degree features. A threshold is used to remove the non-vehicle initial candidates. Experimental results show that this method can adapt to different light conditions robustly with a detection rate over 92%. The proposed method is better than traditional methods based on learning with a higher detection rate and lower error rate.

[ 可能符合您检索需要的词汇 ]

进入期刊文章
针对PID控制中的参数整定的难点及基本BP算法收敛速度慢、易陷入局部极值的问题,提出利用PSO算法的全局寻优能力和较强的收敛性来改进BP网络的权值调整新方法,从而对PID控制的比例、积分、微分进行优化控制。该方法是在基本BP算法的误差反向传播的基础上,使粒子位置的更新对应BP网络的权值和阈值的调整,既充分利用了PSO算法的全局寻优性又较好地保持了BP算法本身的反向传播特点。仿真结果表明基于PSO算法的BP神经网络的PID优化控制具有较好的性能和自学习、自适应性。
In view of the difficulty of parameters setting of PID control and the limitations of slow convergence and local extreme values of BP algorithm,a new method to adjust weights of BP network is proposed using the global optimization ability and the strong conver-gence by PSO algorithm,so as to optimize the proportional,integral and differential of PID control. The new algorithm is based on the weight adjustments of error back propagation of BP algorithm,making the bats position updating to weight and threshold of BP network modification. The new algorithm can not only use the global optimization of PSO algorithm,but also contain the feature of error back propagation of BP algorithm. Experimental results show that the PID optimization control based on BP neural network has better perform-ance and self learning and adaptive.

[ 可能符合您检索需要的词汇 ]

进入期刊文章
收敛性与鲁棒性是模糊神经网络的两个重要性质。对带阈值的Max-T模糊Hopfield神经网络（记为Max-T-C FHNN）的收敛性及在训练模式小幅摄动情况下的鲁棒性进行了分析，从理论上给出了严格的证明。发现了采用最大权值矩阵学习算法时，Max-T-C FHNN具有良好的收敛性，同时当T模及其蕴含算子满足Lipschitz条件时， Max-T-C FHNN对训练模式摄动全局拥有好的鲁棒性，用自联想实验验证了理论的有效性。
Convergence and robustness are two important properties of fuzyy neural network. This paper analyses the conver-gence and robustness of Max-T fuzzy Hopfield neural network with threshold(called Max-T-C FHNN)in the condition of perturbations of training patterns, which is proved theoretically. It is discovered that Max-T-C FHNN using maximum weight matrix is of excellent convergence. Max-T-C FHNN holds good robustness globally to perturbations of training patterns in the case that T-norms and its implication operator satisfy the Lipschitz condition. The self-association experiment is given to testify the theoretical results.

[ 可能符合您检索需要的词汇 ]

进入期刊文章
文章提出一种模拟退火(SA)与粒子群优化(PSO)算法相结合的算法来优化Elman神经网络权值和阈值。当PSO处于停滞状态时，利用粒子群优化算法的全局寻优性质，以及SA能跳出局部最优解的特性，在搜索到的最优位置处用模拟退火算法继续寻找最优解，并对具有动态递归性能的Elman神经网络进行学习训练，这样就能对忙时话务量进行预测。结果表明，与传统Elman神经网络和PSO-Elman神经网络相比，基于模拟退火粒子群算法训练的神经网络具有更高的预测精度和良好的自适应性。
This paper presents a hybrid algorithm that combines simulated annealing (SA) algorithm with parti-cle swarm optimization (PSO) algorithm to optimize the weights and threshold of Elman neural network. By using the advantages of global optimization of PSO, when it is trapped into local optimum, SA is employed to jump out of local optimal solution to find the global optimal solution. The hybrid algorithm is used to train Elman neural network with dynamic recursive properties. The approach is carried out on the forecasting of the busy telephone traffic. The experimental results show that SAPSO-Elman neural network has better precision and adaptability compared with the traditional neural network.

[ 可能符合您检索需要的词汇 ]