Intel Corporation

Qualcomm Incorporated

Microsoft Technology Licensing, LLC

Verizon Media Inc.

Other

Broadcom Corporation

Yahoo Ad Tech LLC

The Scripps Research Institute

Jun Shi

Animesh Mishra

David Jonathan Julian

Amal Ekbal

Sandeep K Jain

Pochang Hsu

Qinghua Li

David Wyatt

Petru Cristian Budianu

Andrew Feng

Prasanna Chromepet Padmanabhan

Somesh Goel

Xintian Eddie Lin

John B Halbert

Chong U Lee

Erik Ordentlich

Seh Kwa

David Zimmerman

Scott Manchester

Huiji Gao

Na Li

Yizhong Wu

Melik Isbara

Milind Rao

Arya Behzad

Peter J Black

Kuljit Singh Bains

Yanli Zhang

Aaron K Martin

George Vergis

Bill Nale

Sumeet Sandhu

Minnie Ho

Feng Xue

Liang Zhang

Nihar Jindal

Guangjie Li

Bo Long

Akihiro Takagi

Maximino Vasquez

Karl H Mauritz

Mridul Jain

Shanshan Zheng

Hongmei Sun

Ling Su

Haichao Wei

Ravi Ranganathan

Phil Baran

Peter Cnudde

Yan Zhang

Abhimanyu Lad

Marcellus Forbes

Yu Liu

Peter John Black

Sriram Sundararajan

Carlos A Guerrero

Thomas Dean Skelton

Robert J Brennan

Ryan Shenvi

Petru Christian Budianu

Christian Cruz Montoya

Michaeel M Kazi

Yu Gan

Gonzalo Jorge Aniano Porcile

Xuling Luo

Rongbin Han

Weijian Wang

Qiang Xiao

Honglu Chen

Pin Sun

Ying Xiong

Xiaowei Liu

Chuang-Chuang Li

Supisa Lerstaveesin

Phil S Baran

Ryan A Shenvi

The present disclosure relates to implementing, updating, and managing operation of a client agent on a client device (e.g., computing device, virtual device) in a way that enables isolation of features and functionality while also allowing the client agent to self-heal and intelligently update discrete features thereon. The client agent includes a collection of plugins that are isolated and run in accordance with respective plugin policies. The client agent makes use of device-level, agent-level, and plugin-level health monitors that collectively monitor a health status of discrete components of the client agent in a way that enables the client agent to selectively discontinue scheduling certain plugins without interrupting functionality of other plugins or of the client agent as a whole. Indeed, features described herein enable the client agent to intelligently update and self-heal with respect to individual plugins based on information obtained by the respective health monitors.

Implementing a policy-based agent with plugin infrastructure

In an example, a particular type of deep learning model is used in the global model of the GDMix model: a Factorization Machine. A Factorization Machine combines a Support Vector Machine (SVM) and Matrix Factorizations. It has the advantage of modeling data with huge sparsity well, while maintaining a linear time complexity. A modification may be further made to the Factorization Machine by introducing Lnorm reduction. This acts to divide calculations made by the Factorization Machine into a portion that can be precomputed and a portion that cannot be precomputed. The portion that can be precomputed is then precomputed in an offline manner. As such, when the model is operated in an online manner, the Factorization Machine only needs to compute the portion that cannot be precomputed, reducing the number of operations that need to performed at runtime and greatly improving processing speed over prior machine learned models.

Factorization machine with L-2 norm reduction for machine learned models

A computer implemented method includes granting a subscriber client access to a cloud service-based resource group within a subscriber controlled computing environment. The subscriber controlled computing environment has a virtual network. A direct network connection is established via the virtual network to a domain controller in the subscriber controlled computing environment. The virtual network is extended to a subscriber client virtual machine in a producer cloud service. The subscriber client virtual machine is then joined with the virtual network. Access to the subscriber client virtual machine is provided via the subscriber controlled computing environment and virtual network.

Virtual machine provisioning and directory service management

The present teaching relates to obtaining a model for identifying content matching a query. Training data are received which include queries, advertisements, and hyperlinks. A plurality of subwords are identified from each of the queries and a plurality of vectors for the plurality of subwords of each of the queries are obtained. Via a neural network, a vector for each of the queries is derived based on a plurality of vectors for the plurality of subwords of the query. A query/ads model is obtained via optimization with respect to an objective function, based on vectors associated with the plurality of subwords of each of the queries and vectors for the queries obtained from the neural network.

System and method for query to ad matching using deep neural net based query embedding

A computer implemented method includes granting a tenant administrator client machine access to a cloud hosted tenant service joined to a directory service. A bulk token for the tenant is obtained in response to a request received from the tenant administrator client machine. An identifier of an authorized tenant client to the cloud hosted tenant service is received and results in the provisioning of a tenant client virtual machine in a cloud service for the authorized tenant client in accordance with a specified provisioning package associated with the bulk token. The tenant client virtual machine is then joined to the directory service. On receipt of an authorized client token at the cloud hosted tenant service from a tenant client machine, the tenant client machine is provided a connection to the tenant client virtual machine.

The present teaching relates to obtaining a model for identifying content matching a query. Training data are received which include queries, advertisements, and hyperlinks. A plurality of subwords are identified from each of the queries and a plurality of input vectors for the plurality of subwords of each of the queries are obtained and an input vector for each of the queries is derived based on a plurality of input vectors of a plurality of subwords of the query. A query/ads model is optimized with respect to an objective function via training an input vector for each of the plurality of subwords associated with each of the queries, an input vector for each of the advertisements and links, and a matrix.

System and method for query to ad matching via optimizing input vectors of query subwords

The disclosed embodiments provide a system for performing a natural language search. During operation, the system applies a first machine learning model to a natural language query to predict one or more search intentions associated with the natural language query. Next, the system applies a second machine learning model to the natural language query to produce one or more search parameters associated with a first intention in the search intention(s), wherein the search parameter(s) include a field and a value of the field. The system then performs a first search of a first vertical associated with the first intention using the search parameter(s). Finally, the system generates a ranking containing a first set of search results from the first search of the first vertical and outputs at least a portion of the ranking in a response to the natural language query.

Vertical processing of natural language searches

The present teaching relates to distributed deep machine learning on a cluster. In one example, a request is received for estimating one or more parameters associated with a machine learning model on a cluster including a plurality of nodes. A set of data is obtained to be used for estimating the one or more parameters. The set of data is divided into a plurality of sub-sets of data, each of which corresponds to one of the plurality of nodes. Each sub-set of data is allocated to a corresponding node for estimating values of the one or more parameters based on the sub-set of data. Estimated values of the one or more parameters obtained based on a corresponding sub-set of data allocated to the node, are received from each of the plurality of nodes. The one or more parameters of the machine learning model are estimated based on the estimated values of the one or more parameters generated by at least some of the plurality of nodes.

Method and system for distributed deep machine learning

Systems and methods of managing memory devices provide for reduced power consumption and better thermal management through enhanced memory throttling. In one embodiment a memory unit includes a memory device and a temperature measurement module coupled to the memory device. The temperature measurement device measures the internal temperature of the memory device. Memory throttling can therefore be implemented based on more accurate measurements and with a much shorter response time.

Throttling memory in response to an internal temperature of a memory device

In some embodiments one or more areas of interest to a user of a display screen are determined. Portions of the display screen other than the one or more areas of interest are dimmed. Other embodiments are described and claimed.

Power savings for display panels

A communication device, such as a smart phone, includes transmit/receive logic to cancel an interfering signal component. The interfering signal component may originate from a communication interface on the device itself. For example, transmissions from the communication interface may interfere with received signals at other communication interfaces on the device. Transmit/receive logic on others of the communication interfaces may use known characteristics of the interfering signal component to cancel the interfering signal component.

Self-interference cancellation

System and method are disclosed for synchronization of a transmitting device and a receiving device that communicate with each other via pulse modulation. The synchronization technique entails the transmitting device sending one or more quasi-periodic pulse sequences to the receiving device. A quasi-periodic pulse sequence is based on a substantially periodic pulse sequence, and may include some non-periodic pulses or not include some periodic pulses. The transmitting device may transmit frames each including a preamble that comprises one or more quasi-periodic pulse sequences, and a data payload that may comprise data. The receiving device receives the signal, generates samples of the signal, and detects the quasi-periodic pulse sequences in the received signal by analyzing samples based on a sample associated with a pulse and the period associated with the substantially periodic pulse sequence. The receiving device is further able to detect frames based on the detection of the sequence, and extract data therefrom.

System and method for transmitting and receiving signal with quasi-periodic pulse sequence

Throttling memory in a computer system

Techniques are described to transmit commands to a display device during vertical or horizontal blanking intervals. The commands can be transmitted using fields that would otherwise be used to transmit color information. A Low Voltage Differential Signaling (LVDS) compliant interface can be used to transmit the commands.

Techniques to control display activity

Growing community of inventors

San Jose, CA, United States of America

Jun Shi