公开(公告)号：US09514404B1

公开(公告)日：2016-12-06

申请号：US14860497

申请日：2015-09-21

Applicant: Google Inc.

Inventor： Gregory S. Corrado ,  Kai Chen ,  Jeffrey A. Dean ,  Gary R. Holt ,  Julian P. Grady ,  Sharat Chikkerur ,  David W. Sculley, II

IPC: G06F15/18 ,  G06E1/00 ,  G06E3/00 ,  G06G7/00 ,  G06N3/08 ,  G06N3/04

CPC classification number: G06N3/08 ,  G06N3/04 ,  G06N3/0454 ,  G06N3/084

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for using embedded function with a deep network. One of the methods includes receiving an input comprising a plurality of features, wherein each of the features is of a different feature type; processing each of the features using a respective embedding function to generate one or more numeric values, wherein each of the embedding functions operates independently of each other embedding function, and wherein each of the embedding functions is used for features of a respective feature type; processing the numeric values using a deep network to generate a first alternative representation of the input, wherein the deep network is a machine learning model composed of a plurality of levels of non-linear operations; and processing the first alternative representation of the input using a logistic regression classifier to predict a label for the input.

公开(公告)号：US09477656B1

公开(公告)日：2016-10-25

申请号：US14084319

申请日：2013-11-19

Applicant: Google Inc.

Inventor： Jeffrey A. Dean

IPC: G06F17/28

CPC classification number: G06F17/289 ,  G06F17/2845 ,  G06F17/30401

Abstract: Systems and methods are disclosed for searching across multi-lingual information. A user makes a query in a first language, and a group of documents that were previously machine-translated into the first language are searched for information responsive to the query. Contextual information derived can be used to improve the accuracy of the machine translation. Responsive documents are returned to the user. Alternatively, a query provided in a user's language may be translated into one or more other languages. Documents written in these languages can then be searched for information responsive to the appropriate translated query. Responsive documents can be translated into the user's language prior to providing them to the user.

公开(公告)号：US20160012331A1

公开(公告)日：2016-01-14

申请号：US14860462

申请日：2015-09-21

Applicant: Google Inc.

Inventor： Kai Chen ,  Xiaodan Song ,  Gregory S. Corrado ,  Kun Zhang ,  Jeffrey A. Dean ,  Bahman Rabii

IPC: G06N3/08 ,  G06F17/30 ,  G06N3/04

CPC classification number: G06N3/084 ,  G06F17/30707 ,  G06F17/30864 ,  G06N3/04 ,  G06N3/0427 ,  G06N3/08 ,  G06Q30/02

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for scoring concept terms using a deep network. One of the methods includes receiving an input comprising a plurality of features of a resource, wherein each feature is a value of a respective attribute of the resource; processing each of the features using a respective embedding function to generate one or more numeric values; processing the numeric values to generate an alternative representation of the features of the resource, wherein processing the floating point values comprises applying one or more non-linear transformations to the floating point values; and processing the alternative representation of the input to generate a respective relevance score for each concept term in a pre-determined set of concept terms, wherein each of the respective relevance scores measures a predicted relevance of the corresponding concept term to the resource.

Abstract translation: 方法，系统和装置，包括在计算机存储介质上编码的计算机程序，用于使用深层网络评分概念术语。 所述方法之一包括接收包括资源的多个特征的输入，其中每个特征是所述资源的相应属性的值; 使用相应的嵌入功能处理每个特征以生成一个或多个数值; 处理所述数值以产生所述资源的特征的替代表示，其中处理所述浮点值包括将一个或多个非线性变换应用于所述浮点值; 以及处理所述输入的替代表示，以在预定概念术语集中为每个概念项产生相应的相关性得分，其中各个相关性分数中的每一个测量相应概念项与资源的预测相关性。

公开(公告)号：US09218573B1

公开(公告)日：2015-12-22

申请号：US13826327

申请日：2013-03-14

Applicant: Google Inc.

Inventor： Gregory S. Corrado ,  Kai Chen ,  Jeffrey A. Dean ,  Samy Bengio ,  Rajat Monga ,  Matthieu Devin

IPC: G06F15/18 ,  G06N99/00 ,  G06N7/00 ,  G06N5/02

CPC classification number: G06N99/005 ,  G06K9/6256 ,  G06K9/6269 ,  G06N3/063 ,  G06N3/08 ,  G06N5/025 ,  G06N7/005 ,  G06N7/08

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for training a model using parameter server shards. One of the methods includes receiving, at a parameter server shard configured to maintain values of a disjoint partition of the parameters of the model, a succession of respective requests for parameter values from each of a plurality of replicas of the model; in response to each request, downloading a current value of each requested parameter to the replica from which the request was received; receiving a succession of uploads, each upload including respective delta values for each of the parameters in the partition maintained by the shard; and updating values of the parameters in the partition maintained by the parameter server shard repeatedly based on the uploads of delta values to generate current parameter values.

Abstract translation: 方法，系统和装置，包括在计算机存储介质上编码的计算机程序，用于使用参数服务器分片训练模型。 其中一种方法包括在被配置为维持模型的参数的不相交分区的值的参数服务器分片上接收来自模型的多个副本中的每一个的参数值的相继请求; 响应于每个请求，将每个请求的参数的当前值下载到从其接收请求的副本; 接收连续的上传，每次上传包括由分片保存的分区中的每个参数的各自的增量值; 并且根据增量值的上载重复地更新由参数服务器分片保存的分区中的参数的值，以生成当前参数值。

公开(公告)号：US08768870B1

公开(公告)日：2014-07-01

申请号：US13968019

申请日：2013-08-15

Applicant: Google Inc.

Inventor： Gregory S. Corrado ,  Kai Chen ,  Jeffrey A. Dean ,  Samy Bengio ,  Rajat Monga ,  Matthieu Devin

IPC: G06F15/18

CPC classification number: G06N99/005 ,  G06K9/6256 ,  G06K9/6269 ,  G06N3/063 ,  G06N3/08 ,  G06N5/025 ,  G06N7/005 ,  G06N7/08

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for training a model using parameter server shards. One of the methods includes receiving, at a parameter server shard configured to maintain values of a disjoint partition of the parameters of the model, a succession of respective requests for parameter values from each of a plurality of replicas of the model; in response to each request, downloading a current value of each requested parameter to the replica from which the request was received; receiving a succession of uploads, each upload including respective delta values for each of the parameters in the partition maintained by the shard; and updating values of the parameters in the partition maintained by the parameter server shard repeatedly based on the uploads of delta values to generate current parameter values.

Abstract translation: 方法，系统和装置，包括在计算机存储介质上编码的计算机程序，用于使用参数服务器分片训练模型。 其中一种方法包括在被配置为维持模型的参数的不相交分区的值的参数服务器分片上接收来自模型的多个副本中的每一个的参数值的相继请求; 响应于每个请求，将每个请求的参数的当前值下载到从其接收请求的副本; 接收连续的上传，每次上传包括由分片保存的分区中的每个参数的各自的增量值; 并且根据增量值的上载重复地更新由参数服务器分片保存的分区中的参数的值，以生成当前参数值。

公开(公告)号：US20140040027A1

公开(公告)日：2014-02-06

申请号：US13960515

申请日：2013-08-06

Applicant: Google Inc.

Inventor： Darrell Anderson ,  Paul Buchheit ,  Alexander Paul Carobus ,  Yingwei Cui ,  Jeffrey A. Dean ,  Georges R. Harik ,  Deepak Jindal ,  Narayanan Shivakumar

IPC: G06Q30/02

CPC classification number: G06F17/30864 ,  G06F17/30454 ,  G06Q30/02 ,  G06Q30/0251 ,  Y10S707/944 ,  Y10S707/99945 ,  Y10S707/99948

Abstract: Advertisers are permitted to put targeted ads on page on the web (or some other document of any media type). The present invention may do so by (i) obtaining content that includes available spots for ads, (ii) determining ads relevant to content, and/or (iii) combining content with ads determined to be relevant to the content.

Abstract translation: 允许广告商在网页上放置有针对性的广告（或任何媒体类型的其他文档）。 本发明可以通过（i）获得包括广告的可用点的内容，（ii）确定与内容相关的广告，和/或（iii）将内容与确定为与内容相关的广告组合。

公开(公告)号：US10152520B1

公开(公告)日：2018-12-11

申请号：US15085826

申请日：2016-03-30

Applicant: Google Inc.

Inventor： Jeffrey A. Dean ,  Corin Anderson ,  Alexis Battle

IPC: G06F17/30 ,  G06F17/22

Abstract: A system generates a model based on feature data relating to different features of a link from a linking document to a linked document and user behavior data relating to navigational actions associated with the link. The system also assigns a rank to a document based on the model.

公开(公告)号：US09477714B1

公开(公告)日：2016-10-25

申请号：US14488795

申请日：2014-09-17

Applicant: Google Inc.

Inventor： Krishna Bharat ,  Jeffrey A. Dean ,  Michael Curtiss ,  Amitabh Singhal ,  Michael Schmitt

IPC: G06F17/30

CPC classification number: G06F17/3053 ,  G06F17/30011 ,  G06F17/30598 ,  G06F17/30675 ,  G06F17/3071 ,  G06F17/30867

Abstract: Methods and apparatus are described for scoring documents in response, in part, to parameters related to the document, source, and/or cluster score. Methods and apparatus are also described for scoring a cluster in response, in part, to parameters related to documents within the cluster and/or sources corresponding to the documents within the cluster. In one embodiment, the invention may detect at least one document within the cluster; analyze a parameter corresponding to the document; and compute a cluster score based, in part, on the parameter, wherein the cluster score corresponds with at least one document within the cluster.

Abstract translation: 描述了对文档进行评分的方法和装置，部分地响应于与文档，源和/或聚类分数相关的参数。 还描述了用于对集群进行评分的方法和装置，部分地涉及与集群内的文档相关的参数和/或对应于集群内的文档的源。 在一个实施例中，本发明可以检测群集内的至少一个文档; 分析与文档相对应的参数; 并且部分地基于所述参数来计算聚类分数，其中所述聚类分数对应于所述聚类内的至少一个文档。

公开(公告)号：US09298852B2

公开(公告)日：2016-03-29

申请号：US13928868

申请日：2013-06-27

Applicant: GOOGLE INC.

Inventor： Emanuel Taropa ,  Jeffrey A. Dean ,  Simon Tong ,  Nitin Gupta ,  Noam M. Shazeer ,  Anwis Das ,  Murtaza A. Basrai

IPC: G06F17/30 ,  G06F7/00

CPC classification number: G06F17/3097 ,  G06F17/3064 ,  G06F17/30646 ,  G06F17/30672

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for reranking query completions based on activity session data. One of the methods includes receiving a query prefix from a user. Query completions are obtained for the query prefix. One or more likely queries that are likely to co-occur with a reference query in user activity sessions are obtained. If one of the likely queries matches one of the query completions, a modified ranking of the query completions is determined, including boosting a ranking of matching query completions. The modified ranking of the query completions is provided in response to receiving the query prefix.

Abstract translation: 方法，系统和装置，包括在计算机存储介质上编码的计算机程序，用于基于活动会话数据重新排列查询完成。 其中一种方法包括从用户接收查询前缀。 获取查询前缀的查询完成。 获得在用户活动会话中可能与参考查询共存的一个或多个可能的查询。 如果一个可能的查询与其中一个查询完成相匹配，则确定查询完成的修改排名，包括提升匹配查询完成的排名。 响应于接收查询前缀而提供查询完成的修改排名。

公开(公告)号：US20150356461A1

公开(公告)日：2015-12-10

申请号：US14731349

申请日：2015-06-04

Applicant: Google Inc.

Inventor： Oriol Vinyals ,  Jeffrey A. Dean ,  Geoffrey E. Hinton

IPC: G06N99/00 ,  G06N7/00

CPC classification number: G06N99/005 ,  G06N3/0454 ,  G06N7/00 ,  G06N7/005

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for training a distilled machine learning model. One of the methods includes training a cumbersome machine learning model, wherein the cumbersome machine learning model is configured to receive an input and generate a respective score for each of a plurality of classes; and training a distilled machine learning model on a plurality of training inputs, wherein the distilled machine learning model is also configured to receive inputs and generate scores for the plurality of classes, comprising: processing each training input using the cumbersome machine learning model to generate a cumbersome target soft output for the training input; and training the distilled machine learning model to, for each of the training inputs, generate a soft output that matches the cumbersome target soft output for the training input.

Abstract translation: 方法，系统和装置，包括在计算机存储介质上编码的计算机程序，用于训练蒸馏机器学习模型。 其中一种方法包括训练繁琐的机器学习模型，其中笨重的机器学习模型被配置为接收输入并为多个类中的每一个生成相应的分数; 并且在多个训练输入上训练蒸馏机器学习模型，其中所述蒸馏机器学习模型还被配置为接收所述多个类别的输入并生成分数，其包括：使用所述麻烦的机器学习模型来处理每个训练输入以产生 训练输入的麻烦目标软输出; 并训练蒸馏机器学习模型，对于每个训练输入，产生与训练输入的麻烦的目标软输出相匹配的软输出。