Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for classifying images.

Abstract: An advertiser specifies a conversion-based bid for a conversion event associated with an ad. If a conversion event occurs for the ad, an effective conversion-based bid can be adjusted by a risk premium associated with the ad. An account associated with the advertiser can be debited based upon the adjusted effective conversion-based bid.

Abstract: An apparatus is described. The apparatus includes a smart image sensor having a memory and a processor that are locally integrated with an image sensor. The memory is to store first program code to be executed by the processor. The memory is coupled to the image sensor and the processor. The memory is to store second program code to be executed by the processor. The first program code is to cause the smart image sensor to perform an analysis on one or more images captured by the image sensor. The analysis identifies a region of interest within the one or more images with machine learning from previously captured images. The second program code is to cause the smart image sensor to change an image sensing and/or optical parameter in response to the analysis of the one or more images performed by the execution of the first program code. Alternatively or in combination, the memory is to store third program code to be executed by the processor and fourth program code to be executed by the processor.

Abstract: A data center cooling system includes an outer container that defines a first volume; an inner container that defines a second volume and is positioned within the first volume, the inner container including an air outlet that includes an airflow path between the first and second volumes; a liquid seal to fluidly isolate a liquid phase of a non-conductive coolant that fills at least a portion of the first and second volumes from an ambient environment; and a plurality of electronic heat-generating devices at least partially immersed in the liquid phase of the non-conductive coolant to transfer a heat load to the non-conductive coolant.

Abstract: Systems and methods are described for estimating a camera pose. The estimation may include obtaining a sequence of images including a plurality of image frames of a scene, detecting a first set of feature points in a first image frame, and tracking the first set of feature points in a plurality of subsequent image frames. While continuing to track the first set of feature points, the estimation can include detecting a second set of feature points in a second image frame, tracking the second set of feature points, selecting a first initial camera pose associated with the first image frame and a second initial camera pose associated with the second image frame, determining projection locations, the projection locations based on the first initial camera pose and the second initial camera pose, and comparing the projection locations corresponding to each feature point in the first and second sets of feature points.

Abstract: A system including a processor and a memory storing instructions that, when executed, cause the system to: receive a stream of content including a plurality of content items; retrieve a plurality of related content items; identify a topic for at least one of the plurality of the related content items; rank the topics by relevance to the user and one of the plurality of items; associate the topic to the content item in the stream of content where the topic is ranked as being relevant to the content item; generate a marker for the topic; generate an interface including the marker associated with the content item in the stream of content; and provide the interface for display.

Abstract: Various implementations process a television content stream to detect program boundaries such as the starting point and ending point of the program. In at least some implementations, program boundaries such as intermediate points between the starting point and ending point of the program are also detected. The intermediate points correspond to where a program pauses for secondary content such as an advertisement or advertisements, and then resumes once the secondary content has run. Once program boundaries are detected, primary content is isolated by removing secondary content that occurs before the starting point and after the ending point. In at least some implementations, secondary content that occurs between detected intermediate points is also removed. The primary content is then recorded without secondary content that originally comprised part of the original television content stream.

Abstract: A circuit for performing neural network computations for a neural network is described. The circuit includes plurality of neural network layers each including a crossbar arrays. The plurality of crossbar arrays are formed in a common substrate in a stacked configuration. Each crossbar array includes a set of crosspoint devices. A respective electrical property of each of the crosspoint devices is adjustable to represent a weight value that is stored for each respective crosspoint device. A processing unit is configured to adjust the respective electrical properties of each of the crosspoint devices by pre-loading each of the crosspoint devices with a tuning signal. A value of the turning signal for each crosspoint device is a function of the weight value represented by each respective crosspoint device.

Abstract: Various arrangements for controlling multiple environmental zones are presented. A first zone specific device may be configured to alter an environmental condition of a first environmental zone of the multiple environmental zones. The first zone specific device may include a rechargeable power source for at least partially powering the operation of the first zone specific device. The first zone specific device may include a communication interface for communicating with other devices of the system. Also, a central controller may be present that is configured to communicate with the first zone specific device to determine a power status of the rechargeable power source of the first zone specific device.

Abstract: A method for providing personalized content sharing is disclosed. The method includes: receiving, by a processing device from a first computing device of a first user, a request pertaining to a video; determining affinity scores corresponding to a first plurality of other users that are contacts of the first user, the determining based on affinities of the first plurality of other users for the video; and providing, to the first computing device, a prompt to the first user to share the video with at least one second user of the first plurality of other users, the at least one second user determined based on the affinity scores.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for gaze position prediction using neural networks. One of the systems includes a neural network comprising one or more neural network layers, wherein the neural network is configured to obtain a collection of input facial images of a user, wherein the collection of input facial images of the user comprises (i) a query image of the user, (ii) one or more calibration images of the user, and (iii) a respective calibration label that labels a known gaze position of the user for each of the one or more calibration images of the user; and process the received collection of input facial images of the user using the one or more neural network layers to generate a neural network output that characterizes a gaze position of the user in the query image.

Abstract: A method for healing a target region on an input image is described. A preview image is received; the preview image may reflect a down-sampled image of an original image. The method determines a target region for the preview image. The target region indicates a segment of the preview image designated for healing. The method may then heal the target region associated with the preview image using a transformation. The method may store one or more parameters associated with the healed preview image. The method may then provide for display the healed preview image to a user on a mobile device.

Abstract: Systems, methods, routines and/or techniques for a local mobile memo for non-interrupting link noting are described. One or more embodiments may include a method that may include displaying to a user a page containing a link, receiving a first user input that indicates the link should be saved for potential later processing and saving the link to a local repository so that the link can be processed later. The saving may occur while the page continues to be displayed. The saving may occur without processing the link to perform the associated action. The method may include displaying a list (e.g., including the link) of one or more links saved in the repository, for example, in response to an event that indicates that the user may be interested in viewing links in the link repository.

Abstract: Methods and devices may be used to remotely hushing alarms by receiving an indication of an alarm event occurring at a remote device and presenting a visual indicator of the alarm event including a hush option to alter an audible and/or visual alarm notification of the remote device. A user selection of the hush option is received. In response to receiving the user selection of the hush option, communicating, from the electronic device to the remote device via a point-to-point low power wireless connection between the electronic device and the remote device, a hush request requests that the remote device alter its audible and/or visual alarm notification and causes an adjustment of the visual indicator to reflect a status of the hush request.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media for learning pronunciations from acoustic sequences. One method includes receiving an acoustic sequence, the acoustic sequence comprising a respective acoustic feature representation at each of a plurality of time steps; for each of the time steps processing the acoustic feature representation through each of one or more recurrent neural network layers to generate a recurrent output; processing the recurrent output for the time step using a phoneme output layer to generate a phoneme representation for the acoustic feature representation for the time step; and processing the recurrent output for the time step using a grapheme output layer to generate a grapheme representation for the acoustic feature representation for the time step; and extracting, from the phoneme and grapheme representations for the acoustic feature representations at each time step, a respective pronunciation for each of one or more words.

Abstract: This document describes a content caching system for pre-loading digital components, the system including a communication interface configured to communicate with a remote device over a wireless network, a local content cache; and an evaluation system comprising one or more processors. The one or more operations include pre-loading a digital component for rendering in a browser at a time that is subsequent to a time of the pre-loading, registering a scheme of a network reference for the cached digital component, with the scheme comprising a specified portion of the network reference for the cached digital component; retrieving, from the local content cache, the pre-loaded digital component associated with the digital component tag comprising the network reference; and rendering, from the local content cache, the pre-loaded digital component in a graphical user interface rather than requesting the digital component from the remote device.

Abstract: Methods, systems, and apparatus include computer programs encoded on a computer-readable storage medium, including a method for providing content. An indication of receipt of a request is received from a user device to navigate to a first web based resource. The request includes an appended cookie when one is available on the requesting device. Partner web based resources are identified in a co-operative that includes the first web based resource. A code segment is provided that is configured to send, to respective entities associated with the identified partner web based resources, a second request to share user data with the first web based resource. The shared user data is data collected by a respective partner web based resource and is associated with the user device. The shared data is data that was collected by an entity associated with the respective partner web based resource and associated with the user device.

Abstract: A method includes performing one or more operations as requested by a thread executing on a processor, the thread having a thread context; receiving a park request from the thread, the park request received following a request from the thread for a low latency resource, wherein the cache response time is less than or equal to a resource response threshold so as to allow the thread context to be stored and retrieved from the cache in less time than the portion of time it takes to complete the request for the low latency resource; storing the thread context in the cache; detecting that the resume condition has occurred; retrieving the thread context from the cache; and resuming execution of the thread.

Abstract: Systems, methods, and media for providing a multipurpose remote control are provided.

Abstract: A virtual view of a scene may be generated through the use of various systems and methods. In one exemplary method, from a tiled array of cameras, image data may be received. The image data may depict a capture volume comprising a scene volume in which a scene is located. A viewing volume may be defined. A virtual occluder may be positioned at least partially within the capture volume such that a virtual window of the virtual occluder is between the viewing volume and the scene. A virtual viewpoint within the viewing volume may be selected. A virtual view may be generated to depict the scene from the virtual viewpoint.