Google Inc.

Microsoft Technology Licensing, LLC

Vmware, Inc.

Apple Inc.

Facebook, Inc.

Waymo LLC

Goolge Inc.

Yang Li

Tsung-Hsiang Chang

Roy Want

William Noah Schilit

Hao Lu

Henry Allan Rowley

Franz Josef Och

Xin Zhou

Gang Li

Shumin Zhai

Lu Xun

John Francis Canny

Zifeng Huang

Yu Ouyang

Mihail Gavril Tarta

Gopal Kakivaya

Christopher Gilliard

Xiangfang Li

James Chi-Shun Tsiao

Konstantin Anatolyevich Bay

Aram Compeau

Si Si

Qiang Wu

Wei W Wang

Sanjiv Kumar

Michael Andrew Cleron

Bay-Wei W Chang

Maria Moon

Gopala Krishna R Kakivaya

Hua-Jun Zeng

Dianne K Hackborn

Shaunak Chatterjee

Xin Liu

Julien Charles Mercay

Ryan Powell

Wei Guang Lu

Lukasz Mieczyslaw Kaiser

Hao-Lin Chiu

Samuel Bengio

Lawrence Chang

Yan Ke

Hans-olav Cavelie

Xiaodong Fan

Anurag Gupta

Mostafa Dehghani

Juliet Rothenberg

Souvik Ghosh

Elin R Pedersen

Lokesh Pooranmal Bajaj

Ed Huai-Hsin Chi

Alexey Alexeevich Gritsenko

Karan Ashok Ahuja

Matthew Snider

Zheng Chen

Nimish Rameshbhai Shah

Cho-Jui Hsieh

Paul Fletcher

Zhiwei Guan

Befekadu Ayenew Ejigou

Ningfeng Liang

Wei Deng

Pei-Hung Chen

Pei-Yu Hou

Pei-Yu Chi

Uran Oh

Matei Negulescu

Leo Chiow-Fuan Ting

Rui Ueyama

Younghun An

Jaime Guillermo Ruiz

Lixing Huang

Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for training and using machine learning models to generate graphical user interfaces from textual descriptions.

Creating user interface using machine learning

A method includes receiving, via a computing device, a screenshot of a display provided by a graphical user interface of the computing device. The method also includes generating, by an image-structure transformer of a neural network, a representation by fusing a first embedding based on the screenshot and a second embedding based on a layout of virtual objects in the screenshot. The method additionally includes predicting, by the neural network and based on the generated representation, a modeling task output associated with the graphical user interface. The method further includes providing, by the computing device, the predicted modeling task output.

Neural networks based multimodal transformer for multi-task user interface modeling

Methods, systems, and apparatus, including computer programs encoded on computer storage media, for implementing an area attention layer in a neural network system. The area attention layer area implements a way for a neural network model to attend to areas in the memory, where each area contains a group of items that are structurally adjacent.

Neural networks with area attention

Techniques for processing application data are described. The techniques include transmitting packets of application data to a data management server. The data management server maintains rules that dictate where the packets go as well as what actions are taken in response to the packets. Based on the rules, the data management server transmits the packets to destinations that may perform various actions. The actions performed by destinations may include logging data received from the applications, or performing other actions such as notifying an administrator or other entity of a particular event, or the like. The described architecture decouples functions such as logging, notification, and the like from the applications for which they are performed, which provides benefits such as improved scalability, as well as removing the need for application developers to build such functionality directly into the applications. This reduces application development time and improves the robustness of application functionality.

Single data transmission using a data management server

A display includes a plurality of pixels grouped into a plurality of lines of pixels. Each line of pixels of the plurality of lines comprises a group of pixels of the plurality of pixels that are coupled to a common scan line as well and that are coupled to different data lines to individually activate each pixel of the group of pixels. The display also includes a common voltage (VCOM) driving circuit configured to receive a waveform and drive the waveform to the display as a VCOM having a value tailored to an individually activated pixel of the group of pixels.

Dynamic VCOM compensation

Aspects of the present disclosure relate to prioritizing notifications for a vehicle operating in an autonomous driving mode. As an example, a set of notifications based on one or more messages received or transmitted by a planning system of the vehicle may be generated. Each notification of the set of notifications is assigned a priority ranking. The set of notifications may be stored in a cache and the one or more processors may determine a first notification for display based on the priority rankings of the set of notifications and the first notification may be displayed on a display of the vehicle to a passenger of the vehicle.

Hierarchical messaging system

An online system obtains a composite prediction associated with a content item indicating a likelihood that a viewing user of the online system will perform a type of conversion associated with the content item via one or more paths of events leading to the type of conversion. The online system obtains the composite prediction based on a tree data structure describing the path(s) of events. Upon identifying an opportunity to present content to the viewing user, the online system identifies the tree data structure corresponding to the type of conversion from multiple tree data structures maintained in the online system and obtains a composite prediction associated with the content item by evaluating and performing operations associated with nodes of the tree data structure while traversing the nodes. The online system then determines whether to present the content item to the viewing user based on the composite prediction.

Obtaining a composite prediction indicating a likelihood that a user of an online system will perform a conversion associated with a content item via one or more paths of events

In an example, first and second machine learned models corresponding to a particular context of a social networking service are obtained, the first machine learned model trained via a first machine learning algorithm to output an indication of importance of a social networking profile field to obtaining results in the particular context, and the second machine learned model trained via a second machine learning algorithm to output a propensity of the user to edit a social networking profile field if requested. One or more missing fields in a social networking profile for the user are identified. For each of one or more of the one or more missing fields, the field and an identification of the user are passed through the first and second machine learned models, and outputs of the first and second machine learned models are combined to identify one or more top missing profile fields.

Machine learned models for contextual editing of social networking profiles

Methods, systems, and apparatus, including computer programs encoded on computer storage media, for decreasing neural network inference times using softmax approximation. One of the methods includes maintaining data specifying a respective softmax weight vector for each output in a vocabulary of possible neural network outputs; receiving a neural network input; processing the neural network input using one or more initial neural network layers to generate a context vector for the neural network input; and generating an approximate score distribution over the vocabulary of possible neural network outputs for the neural network input, comprising: processing the context vector using a screening model configured to predict a proper subset of the vocabulary for the context input; and generating a respective logit for each output that is in the proper subset, comprising applying the softmax weight vector for the output to the context vector.

Decreasing neural network inference times using softmax approximation

In various implementations, methods and systems resource balancing in a distributed computing environment are provided. A client defined resource metric is received that represents a resource of nodes of the cloud computing platform. A placement plan for job instances of service applications is generated. The placement plan includes one or more movements that are executable to achieve a target placement of the job instances on the nodes. It is determined that the placement plan complies with placement rules. Each placement rule dictates whether a given job instance of the job instances is suitable for placement on a given node of the nodes. The placement plan is executed based on determining that the target placement of the job instances improves balance of resources across the nodes of the cloud computing platform based on the resource represented by the client defined resource metric.

Balancing resources in distributed computing environments

Configurable client filtering rules

Methods and apparatus for an interaction system specifying cross-device interaction are provided. The interaction system provides application program interfaces (APIs) and a scripting environment that allow development of scripts that combine user input and sensing events and distribute output and across devices to create a range of rich cross-device behaviors with minimal development. The interaction system includes an integrated environment for developers to author and test cross-device behaviors. When a script is ready, the script can be deployed on a network of mobile and wearable computing devices using the interaction system's runtime environment. An evaluation of the interaction system with twelve participants revealed that the interaction system significantly reduced developer effort for creating and iterating on cross-device behaviors. The interaction system allowed developers to focus on target interaction behaviors and high-level device capabilities, rather than low-level specifications.

Scripting cross-device wearable interaction

A method may include identifying, from a set of applications, a subset of the set of applications, each application from the subset of the set of applications being predicted, by a computing device, to be selected by a user. The method may also include outputting a graphical user interface that includes: a plurality of application icons representing the set of applications and positioned around at least a portion of a perimeter of the graphical user interface; and a plurality of prediction icons positioned within an interior of the graphical user interface and representing the subset of the set of applications. The position of a particular prediction icon representing a particular application may be based on a position of a particular application icon representing the particular application. The method may further include executing an action associated with the particular prediction icon or the one of the plurality of application icons.

Target selection on a small form factor display

Growing community of inventors

Palo Alto, CA, United States of America

Yang Li