The least squares support vector regression (LS-SVR) is usually used for the modeling of single output system, but it is not well suitable for the actual multi-input-multi-output system. The paper aims at the modeling...The least squares support vector regression (LS-SVR) is usually used for the modeling of single output system, but it is not well suitable for the actual multi-input-multi-output system. The paper aims at the modeling of multi-output systems by LS-SVR. The multi-output LS-SVR is derived in detail. To avoid the inversion of large matrix, the recursive algorithm of the parameters is given, which makes the online algorithm of LS-SVR practical. Since the computing time increases with the number of training samples, the sparseness is studied based on the pro-jection of online LS-SVR. The residual of projection less than a threshold is omitted, so that a lot of samples are kept out of the training set and the sparseness is obtained. The standard LS-SVR, nonsparse online LS-SVR and sparse online LS-SVR with different threshold are used for modeling the isomerization of C8 aromatics. The root-mean-square-error (RMSE), number of support vectors and running time of three algorithms are compared and the result indicates that the performance of sparse online LS-SVR is more favorable.展开更多
为实现钢铁件渗碳层深度的在线电磁无损检测,提出在线最小二乘支持向量机(Online Least Square Support Vector Machine,Online LS-SVM)的建模方法。Online LS-SVM是以增量学习训练SVM,以减量学习减少样本数,实现小样本估计的训练方法...为实现钢铁件渗碳层深度的在线电磁无损检测,提出在线最小二乘支持向量机(Online Least Square Support Vector Machine,Online LS-SVM)的建模方法。Online LS-SVM是以增量学习训练SVM,以减量学习减少样本数,实现小样本估计的训练方法。实验结果表明,Online LS-SVM不仅能实现钢铁件渗碳层深度的在线电磁无损检测,而且具有学习速度快、泛化性能好和对样本依赖程度低的优点。展开更多
An online algorithm for training LS-SVM (Least Square Support VectorMachines) was proposed for the application of function estimation and classification. Online LS-SVMmeans that LS-SVM can be trained in an incremental...An online algorithm for training LS-SVM (Least Square Support VectorMachines) was proposed for the application of function estimation and classification. Online LS-SVMmeans that LS-SVM can be trained in an incremental way, and can be pruned to get sparseapproximation in a decremental way. When a SV (Support Vector) is added or removed, the onlinealgorithm avoids computing large-scale matrix inverse. Thus the computation cost is reduced. Onlinealgorithm is especially useful to realistic function estimation problem such as systemidentification. The experiments with benchmark function estimation problem and classificationproblem show the validity of this online algorithm.展开更多
The combustion behavior of two single coals and three coal blends in a 300 kW coal-fired furnace under variableoperating conditions was monitored by a flame monitoring system based on image processing and spectral ana...The combustion behavior of two single coals and three coal blends in a 300 kW coal-fired furnace under variableoperating conditions was monitored by a flame monitoring system based on image processing and spectral analysis. A similaritycoefficient was defined to analyze the similarity of combustion behavior between two different coal types. A total of 20 flamefeatures, extracted by the flame monitoring system, were ranked by weights of their importance estimated using ReliefF, a featureselection algorithm. The mean of the infrared signal was found to have by far the highest importance weight among the flamefeatures. Support vector machine （SVM） was used to identify the coal types. The number of flame features used to build the SVMmodel was reduced from 20 to 12 by combining the methods of ReliefF and SVM, and computational precision was guaranteedsimultaneously. A threshold was found for the relationship between the error rate and similarity coefficient, which were positivelycorrelated. The success rate decreased with increasing similarity coefficient. The results obtained demonstrate that the system canachieve the online＂ identification of coal blends in industry.展开更多
In the contemporary world, digital content that is subject to copyright is facing significant challenges against the act of copyright infringement.Billions of dollars are lost annually because of this illegal act. The...In the contemporary world, digital content that is subject to copyright is facing significant challenges against the act of copyright infringement.Billions of dollars are lost annually because of this illegal act. The currentmost effective trend to tackle this problem is believed to be blocking thosewebsites, particularly through affiliated government bodies. To do so, aneffective detection mechanism is a necessary first step. Some researchers haveused various approaches to analyze the possible common features of suspectedpiracy websites. For instance, most of these websites serve online advertisement, which is considered as their main source of revenue. In addition, theseadvertisements have some common attributes that make them unique ascompared to advertisements posted on normal or legitimate websites. Theyusually encompass keywords such as click-words (words that redirect to installmalicious software) and frequently used words in illegal gambling, illegal sexual acts, and so on. This makes them ideal to be used as one of the key featuresin the process of successfully detecting websites involved in the act of copyrightinfringement. Research has been conducted to identify advertisements servedon suspected piracy websites. However, these studies use a static approachthat relies mainly on manual scanning for the aforementioned keywords. Thisbrings with it some limitations, particularly in coping with the dynamic andever-changing behavior of advertisements posted on these websites. Therefore,we propose a technique that can continuously fine-tune itself and is intelligentenough to effectively identify advertisement (Ad) banners extracted fromsuspected piracy websites. We have done this by leveraging the power ofmachine learning algorithms, particularly the support vector machine with theword2vec word-embedding model. After applying the proposed technique to1015 Ad banners collected from 98 suspected piracy websites and 90 normal orlegitimate websites, we were able to successfully identify Ad banners extractedfrom suspected piracy websites with an accuracy of 97%. We present thistechnique with the hope that it will be a useful tool for various effective piracywebsite detection approaches. To our knowledge, this is the first approachthat uses machine learning to identify Ad banners served on suspected piracywebsites.展开更多
为及时辨识集约化水产养殖水质变化趋势、动态调控水质,确保无应激环境下健康养殖,该文提出了基于时序列相似数据的最小二乘支持向量回归机(least squares support vector regression,LSSVR)水质溶解氧在线预测模型。采用特征点分段时...为及时辨识集约化水产养殖水质变化趋势、动态调控水质,确保无应激环境下健康养殖,该文提出了基于时序列相似数据的最小二乘支持向量回归机(least squares support vector regression,LSSVR)水质溶解氧在线预测模型。采用特征点分段时间弯曲距离(feature points segmented time warping distance,FPSTWD)算法对在线采集的时间序列数据进行分段与相似度计算,以缩减规模的子序列数据集对LSSVR模型进行快速训练优化,实现了多个LSSVR子模型在线建模,将预测数据序列与LSSVR子模型的相似度匹配,自适应地选取最佳的子模型作为在线预测模型。应用该模型对集约化河蟹福利养殖水质参数溶解氧浓度进行在线预测,模型评价指标中最大相对误差、平均绝对百分比误差、相对均方根误差和运行时间分别为4.76%、8.18%、5.23%、8.32 s。研究结果表明,与其他预测方法相比,该模型具有较好的综合预测性能,能够满足河蟹福利养殖水质在线预测的实际需求,并为集约化水产养殖水质精准调控提供研究基础。展开更多
基金Supported by the National Creative Research Groups Science Foundation of China (60721062)the National Basic Research Program of China (2007CB714000)
文摘The least squares support vector regression (LS-SVR) is usually used for the modeling of single output system, but it is not well suitable for the actual multi-input-multi-output system. The paper aims at the modeling of multi-output systems by LS-SVR. The multi-output LS-SVR is derived in detail. To avoid the inversion of large matrix, the recursive algorithm of the parameters is given, which makes the online algorithm of LS-SVR practical. Since the computing time increases with the number of training samples, the sparseness is studied based on the pro-jection of online LS-SVR. The residual of projection less than a threshold is omitted, so that a lot of samples are kept out of the training set and the sparseness is obtained. The standard LS-SVR, nonsparse online LS-SVR and sparse online LS-SVR with different threshold are used for modeling the isomerization of C8 aromatics. The root-mean-square-error (RMSE), number of support vectors and running time of three algorithms are compared and the result indicates that the performance of sparse online LS-SVR is more favorable.
基金Supported by the National Creative Research Groups Science Foundation of P.R. China （NCRGSFC： 60421002） and National High Technology Research and Development Program of China （863 Program） （2006AA04 Z182）
文摘为实现钢铁件渗碳层深度的在线电磁无损检测,提出在线最小二乘支持向量机(Online Least Square Support Vector Machine,Online LS-SVM)的建模方法。Online LS-SVM是以增量学习训练SVM,以减量学习减少样本数,实现小样本估计的训练方法。实验结果表明,Online LS-SVM不仅能实现钢铁件渗碳层深度的在线电磁无损检测,而且具有学习速度快、泛化性能好和对样本依赖程度低的优点。
基金This project was financially supported by the National Natural Science Foundation of China (No. 69889050)
文摘An online algorithm for training LS-SVM (Least Square Support VectorMachines) was proposed for the application of function estimation and classification. Online LS-SVMmeans that LS-SVM can be trained in an incremental way, and can be pruned to get sparseapproximation in a decremental way. When a SV (Support Vector) is added or removed, the onlinealgorithm avoids computing large-scale matrix inverse. Thus the computation cost is reduced. Onlinealgorithm is especially useful to realistic function estimation problem such as systemidentification. The experiments with benchmark function estimation problem and classificationproblem show the validity of this online algorithm.
基金supported by the National Basic Research Program(973 Program)of China(No.2015CB251501)
文摘The combustion behavior of two single coals and three coal blends in a 300 kW coal-fired furnace under variableoperating conditions was monitored by a flame monitoring system based on image processing and spectral analysis. A similaritycoefficient was defined to analyze the similarity of combustion behavior between two different coal types. A total of 20 flamefeatures, extracted by the flame monitoring system, were ranked by weights of their importance estimated using ReliefF, a featureselection algorithm. The mean of the infrared signal was found to have by far the highest importance weight among the flamefeatures. Support vector machine （SVM） was used to identify the coal types. The number of flame features used to build the SVMmodel was reduced from 20 to 12 by combining the methods of ReliefF and SVM, and computational precision was guaranteedsimultaneously. A threshold was found for the relationship between the error rate and similarity coefficient, which were positivelycorrelated. The success rate decreased with increasing similarity coefficient. The results obtained demonstrate that the system canachieve the online＂ identification of coal blends in industry.
基金This research project was supported by the Ministry of Culture,Sports,and Tourism(MCST)and the Korea Copyright Commission in 2021(2019-PF-9500).
文摘In the contemporary world, digital content that is subject to copyright is facing significant challenges against the act of copyright infringement.Billions of dollars are lost annually because of this illegal act. The currentmost effective trend to tackle this problem is believed to be blocking thosewebsites, particularly through affiliated government bodies. To do so, aneffective detection mechanism is a necessary first step. Some researchers haveused various approaches to analyze the possible common features of suspectedpiracy websites. For instance, most of these websites serve online advertisement, which is considered as their main source of revenue. In addition, theseadvertisements have some common attributes that make them unique ascompared to advertisements posted on normal or legitimate websites. Theyusually encompass keywords such as click-words (words that redirect to installmalicious software) and frequently used words in illegal gambling, illegal sexual acts, and so on. This makes them ideal to be used as one of the key featuresin the process of successfully detecting websites involved in the act of copyrightinfringement. Research has been conducted to identify advertisements servedon suspected piracy websites. However, these studies use a static approachthat relies mainly on manual scanning for the aforementioned keywords. Thisbrings with it some limitations, particularly in coping with the dynamic andever-changing behavior of advertisements posted on these websites. Therefore,we propose a technique that can continuously fine-tune itself and is intelligentenough to effectively identify advertisement (Ad) banners extracted fromsuspected piracy websites. We have done this by leveraging the power ofmachine learning algorithms, particularly the support vector machine with theword2vec word-embedding model. After applying the proposed technique to1015 Ad banners collected from 98 suspected piracy websites and 90 normal orlegitimate websites, we were able to successfully identify Ad banners extractedfrom suspected piracy websites with an accuracy of 97%. We present thistechnique with the hope that it will be a useful tool for various effective piracywebsite detection approaches. To our knowledge, this is the first approachthat uses machine learning to identify Ad banners served on suspected piracywebsites.
