Google Inc.

Xerox Corporation

Aquatrove Biosciences

Vineet Gupta

Ramanathan V Guha

Vivek Raghunathan

Ramakrishnan Srikant

Andrew M Dai

Kiran S Panesar

Pablo Bellver

Aparna Chennapragada

Amay Nitin Champaneria

Frederick Peter Brewin

Carolyn F Au

Mahesh Keralapura Manjunatha

Karthik Lakshminarayanan

Carl Lischeske

Ravi Kolluri

Simon Tong

John Ogden Lamping

Mizuki McGrath

Benedict A Gomes

Amitabh Singhal

Shivakumar Venkataraman

Gokay Baris Gultekin

Pradnya Karbhari

Abhinav Jalan

Andrew A Berlin

Daniel Gureasko Bobrow

Tad H Hogg

Amit Singhal

Lalitesh Kumar Katragadda

Piyush Prahladka

Philip Michael Long

Vijay A Saraswat

Boris Nikolic

Ben Gomes

Hanie Sedghi

Vibha Gupta

Ramananthan V Guha

Bjorn R Carlson

Mizuki Mcgrath

Methods, systems, and apparatus, including computer programs encoded on computer storage media, for training a convolutional neural network using a regularization scheme. One of the methods includes repeatedly performing the following operations: obtaining a kernel of a particular convolutional layer; applying a Fourier transform to the kernel; generating a decomposition using singular-value decomposition (SVD); generating a regularized diagonal matrix; generating a recomposition; applying an inverse Fourier transform to the recomposition; and training the convolutional neural network on training inputs.

Regularizing the training of convolutional neural networks

Methods and apparatus for providing travel-related information for a location to a user based on activity indications of the user that are related to the location. The location may be determined based on a set of one or more related activity indications and a travel-related score may be determined for the location that is indicative of likelihood that the user has interest in travelling to the location. The user may be provided the travel-related information for the location based on the travel-related score.

Automatic continued search

A system and method for providing preferred language and/or country ordering of search results is described. A search query describing potentially retrievable information provided in a plurality of search result languages and/or countries is received. A search is executed by evaluating the search query against information characteristics maintained in a searchable data repository. At least one preferred language and/or country applicable to search results generated is dynamically determined responsive to the executed search. At least some of the search results are ordered in consideration of the at least one preferred language and/or country.

Ordering of search results based on language and/or country of the search results

Apparatus, systems and methods for predictive query identification for advertisements are disclosed. Candidate query are identified from queries stored in a query log. Relevancy scores for a plurality of web documents are generated, each relevancy score associated with a corresponding web document and being a measure of the relevance of the candidate query to the web document. A web document having an associated relevancy score that exceeds a relevancy threshold is selected. The selected web document is associated with the candidate query.

Query identification and association

Methods, systems, and apparatus, including computer program products, for identifying candidate synonyms of transliterated terms for query expansion. In one aspect, a method includes identifying multiple transliterated terms in a target language. For each transliterated term of the multiple transliterated terms in the target language, the transliterated term is mapped to one or more terms in a source language. For a first transliterated term of the multiple transliterated terms in the target language, one or more second transliterated terms of the multiple transliterated terms in the target language are identified as candidate synonyms of the first transliterated term, where each of the one or more second transliterated terms is mapped to at least one term in the source language that is also mapped from the first transliterated term.

Transliteration for query expansion

A system and method for providing preferred country ordering of search results is described. A search query describing potentially retrievable information provided in a plurality of search result countries is received. A search is executed by evaluating the search query against information characteristics maintained in a searchable data repository. At least one preferred country applicable to search results generated is dynamically determined responsive to the executed search. At least some of the search results are ordered in consideration of the at least one preferred country.

System and method for providing preferred country biasing of search results

Embedded in a transport assembly are arrays of microelectromechanical sensors and actuators for detecting and propelling an object. A controller having defined therein local computational agents and a global controller controls the array of sensors and actuators. The global controller provides global operating constraints to the local computational agents. The global operating constraints are developed using an approximate specification of system behavior based on simplified assumptions of an idealized system as well as limited sensor information aggregated from the array of sensors. The local computational agents compute a desired local actuator response using sensor information from a localized grouping of sensor units. To improve the accuracy of the global operating constraints, the local computational agents reduce differences between a global actuator response, computed using the global operating constraints, and the desired local actuator response. In addition, the local computational agents reduce the correlation among different parts of the transport assembly by reducing differences between actuator responses of neighborhoods of local computational agents.

Distributed control system with global contraints for controlling object motion with smart matter

Qualitative reasoning for physical systems, and techniques for automatic code generation are used to automatically construct digital controllers/device drivers for electro-mechanical systems. Such construction uses models of the system's components described as finite state machines, to form a configuration space. Transitions of the configuration space are labelled as external or internal and a state of the system is identified as a desired state while other states are identified as undesirable. From this configuration space a controller is generated to drive the system to the desired state while avoiding the undesirable states.

Automatic construction of digital controllers/device drivers for

Growing community of inventors

Palo Alto, CA, United States of America

Vineet Gupta