利索-全球专利检索

  1. 登录
  2. 注册

搜索结果

78 条记录 (耗时 0.026 秒)

    Methods for multi-class cost-sensitive learning
    1.
    发明授权
    Methods for multi-class cost-sensitive learning 有权
    多类成本敏感性学习方法

    公开(公告)号：US07558764B2

    公开(公告)日：2009-07-07

    申请号：US11937629

    申请日：2007-11-09

    申请人： Naoki Abe ,  Bianca Zadrozny

    发明人： Naoki Abe ,  Bianca Zadrozny

    IPC分类号： G06F15/18 ,  G06E1/00 ,  G06E3/00 ,  G06G7/00

    CPC分类号： G06N99/005 ,  G06K9/6256 ,  Y10S706/932

    摘要： Methods for multi-class cost-sensitive learning are based on iterative example weighting schemes and solve multi-class cost-sensitive learning problems using a binary classification algorithm. One of the methods works by iteratively applying weighted sampling from an expanded data set, which is obtained by enhancing each example in the original data set with as many data points as there are possible labels for any single instance, using a weighting scheme which gives each labeled example the weight specified as the difference between the average cost on that instance by the averaged hypotheses from the iterations so far and the misclassification cost associated with the label in the labeled example in question. It then calls the component classification algorithm on a modified binary classification problem in which each example is itself already a labeled pair, and its (meta) label is 1 or 0 depending on whether the example weight in the above weighting scheme is positive or negative, respectively. It then finally outputs a classifier hypothesis which is the average of all the hypotheses output in the respective iterations.

    摘要翻译： 多类成本敏感学习的方法基于迭代示例加权方案，并使用二进制分类算法解决多类成本敏感学习问题。 其中一种方法通过迭代地应用来自扩展数据集的加权采样来工作，该扩展数据集通过使用赋予每个实例的加权方案来增强具有尽可能多的数据点的数据点与原始数据集中的每个示例而获得的加权采样 标示的重量指定为该实例的平均成本与目前为止的迭代的平均假设之间的差异以及与所标记的示例中的标签相关联的错误分类成本。 然后，对修改后的二进制分类问题调用组件分类算法，其中每个示例本身已经是一个标记对，根据上述加权方案中的示例权重是正还是负，其（元）标签为1或0， 分别。 然后，它最终输出一个分类器假设，它是相应迭代中输出的所有假设的平均值。

    Methods for multi-class cost-sensitive learning
    2.
    发明申请
    Methods for multi-class cost-sensitive learning 审中-公开
    多类成本敏感性学习方法

    公开(公告)号：US20050289089A1

    公开(公告)日：2005-12-29

    申请号：US10876533

    申请日：2004-06-28

    申请人： Naoki Abe ,  Bianca Zadrozny

    发明人： Naoki Abe ,  Bianca Zadrozny

    IPC分类号： G06F9/44 ,  G06F15/18 ,  G06K9/62 ,  G06N3/08 ,  G06N7/02 ,  G06N7/06

    CPC分类号： G06K9/6256 ,  G06N20/00 ,  Y10S706/932

    摘要： Methods for multi-class cost-sensitive learning are based on iterative example weighting schemes and solve multi-class cost-sensitive learning problems using a binary classification algorithm. One of the methods works by iteratively applying weighted sampling from an expanded data set, which is obtained by enhancing each example in the original data set with as many data points as there are possible labels for any single instance, using a weighting scheme which gives each labeled example the weight specified as the difference between the average cost on that instance by the averaged hypotheses from the iterations so far and the misclassification cost associated with the label in the labeled example in question. It then calls the component classification algorithm on a modified binary classification problem in which each example is itself already a labeled pair, and its (meta) label is 1 or 0 depending on whether the example weight in the above weighting scheme is positive or negative, respectively. It then finally outputs a classifier hypothesis which is the average of all the hypotheses output in the respective iterations.

    摘要翻译： 多类成本敏感学习的方法基于迭代示例加权方案，并使用二进制分类算法解决多类成本敏感学习问题。 其中一种方法通过迭代地应用来自扩展数据集的加权采样来工作，该扩展数据集通过使用给出每个实例的加权方案来增强原始数据集中具有尽可能多的数据点的数据点与任何单个实例的可能标签而获得的每个示例而获得 标示的重量指定为该实例的平均成本与目前为止的迭代的平均假设之间的差异以及与所标记的示例中的标签相关联的错误分类成本。 然后，对修改后的二进制分类问题调用组件分类算法，其中每个示例本身已经是一个标记对，根据上述加权方案中的示例权重是正还是负，其（元）标签为1或0， 分别。 然后，它最终输出一个分类器假设，它是相应迭代中输出的所有假设的平均值。

    METHODS FOR MULTI-CLASS COST-SENSITIVE LEARNING
    3.
    发明申请
    METHODS FOR MULTI-CLASS COST-SENSITIVE LEARNING 有权
    多级成本敏感性学习方法

    公开(公告)号：US20080065572A1

    公开(公告)日：2008-03-13

    申请号：US11937629

    申请日：2007-11-09

    申请人： Naoki Abe ,  Bianca Zadrozny

    发明人： Naoki Abe ,  Bianca Zadrozny

    IPC分类号： G06N3/00

    CPC分类号： G06N99/005 ,  G06K9/6256 ,  Y10S706/932

    摘要： Methods for multi-class cost-sensitive learning are based on iterative example weighting schemes and solve multi-class cost-sensitive learning problems using a binary classification algorithm. One of the methods works by iteratively applying weighted sampling from an expanded data set, which is obtained by enhancing each example in the original data set with as many data points as there are possible labels for any single instance, using a weighting scheme which gives each labeled example the weight specified as the difference between the average cost on that instance by the averaged hypotheses from the iterations so far and the misclassification cost associated with the label in the labeled example in question. It then calls the component classification algorithm on a modified binary classification problem in which each example is itself already a labeled pair, and its (meta) label is 1 or 0 depending on whether the example weight in the above weighting scheme is positive or negative, respectively. It then finally outputs a classifier hypothesis which is the average of all the hypotheses output in the respective iterations.

    摘要翻译： 多类成本敏感学习的方法基于迭代示例加权方案，并使用二进制分类算法解决多类成本敏感学习问题。 其中一种方法通过迭代地应用来自扩展数据集的加权采样来工作，该扩展数据集通过使用给出每个实例的加权方案来增强原始数据集中具有尽可能多的数据点的数据点与任何单个实例的可能标签而获得的每个示例而获得 标示的重量指定为该实例的平均成本与目前为止的迭代的平均假设之间的差异以及与所标记的示例中的标签相关联的错误分类成本。 然后，对修改后的二进制分类问题调用组件分类算法，其中每个示例本身已经是一个标记对，根据上述加权方案中的示例权重是正还是负，其（元）标签为1或0， 分别。 然后，它最终输出一个分类器假设，它是相应迭代中输出的所有假设的平均值。

    Resource-light method and apparatus for outlier detection
    6.
    发明授权
    Resource-light method and apparatus for outlier detection 失效
    资源光法和异常检测装置

    公开(公告)号：US08006157B2

    公开(公告)日：2011-08-23

    申请号：US11863704

    申请日：2007-09-28

    申请人： Naoki Abe ,  John Langford

    发明人： Naoki Abe ,  John Langford

    IPC分类号： H04L1/00 ,  G06F11/30 ,  H03M13/00

    CPC分类号： G06N99/005 ,  Y10S707/99936 ,  Y10S707/99943

    摘要： Outlier detection methods and apparatus have light computational resources requirement, especially on the storage requirement, and yet achieve a state-of-the-art predictive performance. The outlier detection problem is first reduced to that of a classification learning problem, and then selective sampling based on uncertainty of prediction is applied to further reduce the amount of data required for data analysis, resulting in enhanced predictive performance. The reduction to classification essentially consists in using the unlabeled normal data as positive examples, and randomly generated synthesized examples as negative examples. Application of selective sampling makes use of an underlying, arbitrary classification learning algorithm, the data labeled by the above procedure, and proceeds iteratively. Each iteration consisting of selection of a smaller sub-sample from the input data, training of the underlying classification algorithm with the selected data, and storing the classifier output by the classification algorithm. The selection is done by essentially choosing examples that are harder to classify with the classifiers obtained in the preceding iterations. The final output hypothesis is a voting function of the classifiers obtained in the iterations of the above procedure.

    摘要翻译： 异常值检测方法和装置具有较轻的计算资源需求，特别是对存储要求的要求，而且具有最先进的预测性能。 异常值检测问题首先降低到分类学习问题，然后应用基于预测不确定度的选择性抽样来进一步减少数据分析所需的数据量，从而提高预测性能。 归类分类主要在于使用未标记的正常数据作为正例，随机生成合成实例作为阴性实例。 选择性抽样的应用使用了基础的，任意的分类学习算法，由上述过程标记的数据，并且迭代地进行。 每个迭代包括从输入数据中选择较小的子样本，对所选数据训练底层分类算法，以及通过分类算法存储分类器输出。 选择是通过基本上选择难以对上述迭代中获得的分类器进行分类的示例来完成的。 最终输出假设是在上述过程的迭代中获得的分类器的投票函数。

    Light Scanning Device and Image Forming Apparatus
    9.
    发明申请
    Light Scanning Device and Image Forming Apparatus 有权
    光扫描装置及成像装置

    公开(公告)号：US20080266634A1

    公开(公告)日：2008-10-30

    申请号：US12109658

    申请日：2008-04-25

    申请人： Hirokazu Nakashima ,  Naoki Abe

    发明人： Hirokazu Nakashima ,  Naoki Abe

    IPC分类号： G02B26/10 ,  G03G15/22

    CPC分类号： G02B26/105 ,  G02B26/127 ,  G03G15/0435 ,  G03G15/326 ,  G03G2215/0404 ,  H04N1/02815 ,  H04N1/0473 ,  H04N1/113 ,  H04N2201/04703 ,  H04N2201/04734 ,  H04N2201/04748 ,  H04N2201/04755 ,  H04N2201/04767

    摘要： A light scanning device is provided. The light scanning device includes: an oscillating mirror which oscillates rotationally and reflects a light beam to be scanned over a scanning range, the scanning range including a first scanning range and a second scanning range set across a center of the scanning range; a detection unit including a light receiving face, on which the light beam is incident, to detect the light beam; and first and second stationary mirrors which reflect the light beam reflected by the oscillating mirror to the first scanning range and the second scanning range, respectively, to be incident on the light receiving face, wherein an incident pattern of the light beam reflected by the first stationary mirror incident on the light receiving face is different from an incident pattern of the light beam reflected by the second stationary mirror incident on the light receiving face.

    摘要翻译： 提供光扫描装置。 光扫描装置包括：振动反射镜，其在扫描范围内旋转振荡并反射待扫描的光束，扫描范围包括跨越扫描范围的中心设置的第一扫描范围和第二扫描范围; 检测单元，其包括光束入射的光接收面以检测光束; 以及第一和第二固定镜，其将由所述振动反射镜反射的光束分别反射到所述第一扫描范围和所述第二扫描范围，以入射到所述光接收面上，其中所述光束的入射图案被所述第一 入射在光接收面上的固定镜与入射在受光面上的第二静止镜反射的光束的入射图案不同。