利索-全球专利检索

  1. 登录
  2. 注册

搜索结果

8 条记录 (耗时 0.017 秒)

    Parallelization of online learning algorithms
    1.
    发明授权
    Parallelization of online learning algorithms 有权
    在线学习算法的并行化

    公开(公告)号：US08904149B2

    公开(公告)日：2014-12-02

    申请号：US12822918

    申请日：2010-06-24

    申请人： Taha Bekir Eren ,  Oleg Isakov ,  Weizhu Chen ,  Jeffrey Scott Dunn ,  Thomas Ivan Borchert ,  Joaquin Quinonero Candela ,  Thore Kurt Hartwig Graepel ,  Ralf Herbrich

    发明人： Taha Bekir Eren ,  Oleg Isakov ,  Weizhu Chen ,  Jeffrey Scott Dunn ,  Thomas Ivan Borchert ,  Joaquin Quinonero Candela ,  Thore Kurt Hartwig Graepel ,  Ralf Herbrich

    IPC分类号： G06F15/76 ,  G06F9/02 ,  G06N99/00

    CPC分类号： G06N99/005

    摘要： Methods, systems, and media are provided for a dynamic batch strategy utilized in parallelization of online learning algorithms. The dynamic batch strategy provides a merge function on the basis of a threshold level difference between the original model state and an updated model state, rather than according to a constant or pre-determined batch size. The merging includes reading a batch of incoming streaming data, retrieving any missing model beliefs from partner processors, and training on the batch of incoming streaming data. The steps of reading, retrieving, and training are repeated until the measured difference in states exceeds a set threshold level. The measured differences which exceed the threshold level are merged for each of the plurality of processors according to attributes. The merged differences which exceed the threshold level are combined with the original partial model states to obtain an updated global model state.

    摘要翻译： 提供了方法，系统和媒体，用于在线学习算法并行化中使用的动态批处理策略。 动态批量策略基于原始模型状态和更新的模型状态之间的阈值水平差，而不是根据常数或预定的批量大小来提供合并功能。 合并包括读取一批传入的流式传输数据，从合作伙伴处理器中检索任何丢失的模型信念，以及对批量的传入流数据进行培训。 重复读取，检索和训练的步骤，直到测量的状态差异超过设定的阈值水平。 根据属性对多个处理器中的每一个合并超过阈值水平的测量差异。 将超过阈值水平的合并差异与原始部分模型状态相结合以获得更新的全局模型状态。

    Metadata Prediction of Objects in a Social Networking System Using Crowd Sourcing
    2.
    发明申请
    Metadata Prediction of Objects in a Social Networking System Using Crowd Sourcing 有权
    使用人群采购的社交网络系统中对象的元数据预测

    公开(公告)号：US20130151612A1

    公开(公告)日：2013-06-13

    申请号：US13324776

    申请日：2011-12-13

    申请人： Jeffrey Scott Dunn ,  Robert William Cathcart ,  Cameron Alexander Marlow ,  Mukund Narasimhan ,  Yuankai Ge ,  Venkataramanan Iyer Nandagopal ,  Chad Little

    发明人： Jeffrey Scott Dunn ,  Robert William Cathcart ,  Cameron Alexander Marlow ,  Mukund Narasimhan ,  Yuankai Ge ,  Venkataramanan Iyer Nandagopal ,  Chad Little

    IPC分类号： G06F15/16

    CPC分类号： H04L67/306 ,  G06Q10/00 ,  H04L67/22

    摘要： A social networking system leverages user's social information to evaluate content submitted for inclusion in objects. If the evaluated submission is accepted, the submission is added to the content of an object. Accepted submissions are also used to predict associations between metadata and objects. Metadata is used to predict which objects will match user searches for information. The social networking system also provides a user interface configured to prompt users to submit information to objects. When a user completes a submission to an object, the user is provided with other options for groups of objects to contribute to. The objects offered are chosen to increase the likelihood that the user will choose to provide submissions to one of the provided objects.

    摘要翻译： 社交网络系统利用用户的社交信息来评估提交给对象的内容。 如果评估的提交被接受，则将提交添加到对象的内容。 接受的提交也用于预测元数据和对象之间的关联。 元数据用于预测哪些对象将匹配用户搜索的信息。 社交网络系统还提供用户界面，用于提示用户将信息提交给对象。 当用户完成对对象的提交时，向用户提供用于贡献的对象组的其他选项。 所提供的对象被选择以增加用户将选择向所提供的对象之一提供提交的可能性。

    Parallelization of Online Learning Algorithms
    4.
    发明申请
    Parallelization of Online Learning Algorithms 有权
    在线学习算法的并行化

    公开(公告)号：US20110320767A1

    公开(公告)日：2011-12-29

    申请号：US12822918

    申请日：2010-06-24

    申请人： Taha Bekir Eren ,  Oleg Isakov ,  Weizhu Chen ,  Jeffrey Scott Dunn ,  Thomas Ivan Borchert ,  Joaquin Quinonero Candela ,  Thore Kurt Hartwig Graepel ,  Ralf Herbrich

    发明人： Taha Bekir Eren ,  Oleg Isakov ,  Weizhu Chen ,  Jeffrey Scott Dunn ,  Thomas Ivan Borchert ,  Joaquin Quinonero Candela ,  Thore Kurt Hartwig Graepel ,  Ralf Herbrich

    IPC分类号： G06F15/76 ,  G06F9/02

    CPC分类号： G06N99/005

    摘要： Methods, systems, and media are provided for a dynamic batch strategy utilized in parallelization of online learning algorithms. The dynamic batch strategy provides a merge function on the basis of a threshold level difference between the original model state and an updated model state, rather than according to a constant or pre-determined batch size. The merging includes reading a batch of incoming streaming data, retrieving any missing model beliefs from partner processors, and training on the batch of incoming streaming data. The steps of reading, retrieving, and training are repeated until the measured difference in states exceeds a set threshold level. The measured differences which exceed the threshold level are merged for each of the plurality of processors according to attributes. The merged differences which exceed the threshold level are combined with the original partial model states to obtain an updated global model state.

    摘要翻译： 提供了方法，系统和媒体，用于在线学习算法并行化中使用的动态批处理策略。 动态批量策略基于原始模型状态和更新的模型状态之间的阈值水平差，而不是根据常数或预定的批量大小来提供合并功能。 合并包括读取一批传入的流式传输数据，从合作伙伴处理器中检索任何丢失的模型信念，以及对批量的传入流数据进行培训。 重复读取，检索和训练的步骤，直到测量的状态差异超过设定的阈值水平。 根据属性对多个处理器中的每一个合并超过阈值水平的测量差异。 将超过阈值水平的合并差异与原始部分模型状态相结合以获得更新的全局模型状态。

    Identifying relevant data for pages in a social networking system
    7.
    发明授权
    Identifying relevant data for pages in a social networking system 有权
    识别社交网络系统中页面的相关数据

    公开(公告)号：US08935299B2

    公开(公告)日：2015-01-13

    申请号：US13553761

    申请日：2012-07-19

    申请人： Cameron Alexander Marlow ,  Robert William Cathcart ,  Jeffrey Scott Dunn ,  Rafael Linden Sagula ,  Venkataramanan Iyer Nandagopal ,  Siddharth Kar ,  Jonathan Chang

    发明人： Cameron Alexander Marlow ,  Robert William Cathcart ,  Jeffrey Scott Dunn ,  Rafael Linden Sagula ,  Venkataramanan Iyer Nandagopal ,  Siddharth Kar ,  Jonathan Chang

    IPC分类号： G06F7/00 ,  G06F17/30

    CPC分类号： G06F17/30699 ,  G06F17/30716 ,  G06F17/30867

    摘要： Embodiments of the invention improve the ability of a social networking system to determine which types of data—hereinafter referred to as “fields”—are relevant to which types of user pages. Specifically, a social networking system assigns page types to different user pages, and likewise stores information on different types of fields. By analyzing the relationships of different pages and fields, the social networking system determines which types of fields are particularly well-suited for inclusion on different types of pages. Using the learned information about page types and field types, the social networking system can better aid page administrators in specifying data to add to their pages. For example, the social networking system can recommend to administrators the addition of certain types of fields or automatically add the fields. Further, the social networking system can specialize a search for social networking system data to field types.

    摘要翻译： 本发明的实施例提高了社交网络系统确定哪些类型的数据（以下称为“字段”）与哪些类型的用户页面相关的能力。 具体地，社交网络系统将页面类型分配给不同的用户页面，并且同样存储关于不同类型的字段的信息。 通过分析不同页面和领域的关系，社交网络系统确定哪些类型的字段特别适合纳入不同类型的页面。 使用有关页面类型和字段类型的学习信息，社交网络系统可以更好地帮助页面管理员指定要添加到其页面的数据。 例如，社交网络系统可以向管理员推荐添加某些类型的字段或自动添加字段。 此外，社交网络系统可以专门搜索社交网络系统数据到字段类型。

    IDENTIFYING RELEVANT DATA FOR PAGES IN A SOCIAL NETWORKING SYSTEM
    8.
    发明申请
    IDENTIFYING RELEVANT DATA FOR PAGES IN A SOCIAL NETWORKING SYSTEM 有权
    在社交网络系统中识别页面的相关数据

    公开(公告)号：US20140025666A1

    公开(公告)日：2014-01-23

    申请号：US13553761

    申请日：2012-07-19

    申请人： Cameron Alexander Marlow ,  Robert William Cathcart ,  Jeffrey Scott Dunn ,  Rafael Linden Sagula ,  Venkataramanan Iyer Nadagopal ,  Siddharth Kar ,  Jonathan Chang

    发明人： Cameron Alexander Marlow ,  Robert William Cathcart ,  Jeffrey Scott Dunn ,  Rafael Linden Sagula ,  Venkataramanan Iyer Nadagopal ,  Siddharth Kar ,  Jonathan Chang

    IPC分类号： G06F15/16 ,  G06F17/30

    CPC分类号： G06F17/30699 ,  G06F17/30716 ,  G06F17/30867

    摘要： Embodiments of the invention improve the ability of a social networking system to determine which types of data—hereinafter referred to as “fields”—are relevant to which types of user pages. Specifically, a social networking system assigns page types to different user pages, and likewise stores information on different types of fields. By analyzing the relationships of different pages and fields, the social networking system determines which types of fields are particularly well-suited for inclusion on different types of pages. Using the learned information about page types and field types, the social networking system can better aid page administrators in specifying data to add to their pages. For example, the social networking system can recommend to administrators the addition of certain types of fields or automatically add the fields. Further, the social networking system can specialize a search for social networking system data to field types.

    摘要翻译： 本发明的实施例提高了社交网络系统确定哪些类型的数据（以下称为“字段”）与哪些类型的用户页面相关的能力。 具体地，社交网络系统将页面类型分配给不同的用户页面，并且同样存储关于不同类型的字段的信息。 通过分析不同页面和领域的关系，社交网络系统确定哪些类型的字段特别适合纳入不同类型的页面。 使用有关页面类型和字段类型的学习信息，社交网络系统可以更好地帮助页面管理员指定要添加到其页面的数据。 例如，社交网络系统可以向管理员推荐添加某些类型的字段或自动添加字段。 此外，社交网络系统可以专门搜索社交网络系统数据到字段类型。