Abstract: Aspects of the subject technology relate to a method for using a voice command for multiple computing devices. First voice input data is received from a first computing device associated with a user account, where the first voice input data comprises a first voice command captured at the first computing device. Second voice input data is received from a second computing device associated with the user account where the second voice input data comprises a second voice command captured at the second computing device. An intended voice command is determined based on the obtained first and second voice input data. Based on the intended voice command, a first target computing device is determined. First instructions associated with the intended voice command are provided to the first target computing device for execution.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for speech recognition using complex linear projection are disclosed. In one aspect, a method includes the actions of receiving audio data corresponding to an utterance. The method further includes generating frequency domain data using the audio data. The method further includes processing the frequency domain data using complex linear projection. The method further includes providing the processed frequency domain data to a neural network trained as an acoustic model. The method further includes generating a transcription for the utterance that is determined based at least on output that the neural network provides in response to receiving the processed frequency domain data.

Abstract: A method of optimizing sleep of a subject using smart-home devices may include operating a smart-home system that is configured to operate in a normal mode and a sleep mode. The method may also include determining that the smart-home system should transition into the sleep mode. The smart-home devices may use a set of default parameters when operating in the sleep mode. The method may additionally include monitoring, while in the sleep mode, a sleep cycle of the subject using the smart-home devices. The method may further include detecting behavior of the subject that indicates that the sleep cycle of the subject is being interrupted or about to be interrupted, determining an environmental control that corresponds with the behavior of the subject, and adjusting the environmental control using the smart-home devices to prevent or stop the sleep cycle of the subject from being interrupted.

Abstract: Methods, and systems, including computer programs encoded on computer storage media for compressing data items with variable compression rate. A system includes an encoder sub-network configured to receive a system input image and to generate an encoded representation of the system input image, the encoder sub-network including a first stack of neural network layers including one or more LSTM neural network layers and one or more non-LSTM neural network layers, the first stack configured to, at each of a plurality of time steps, receive an input image for the time step that is derived from the system input image and generate a corresponding first stack output, and a binarizing neural network layer configured to receive a first stack output as input and generate a corresponding binarized output.

Abstract: Systems and methods for performing connectivity verification testing and topology discovery in a reconfigurable optical add/drop multiplexer (ROADM) are provided. The ROADM can include a ROADM block having a plurality of internal ports connected to a fiber shuffle via respective optical fibers. The ROADM block includes a test signal transmitter configured to inject an outgoing test signal having a unique signature into each internal port. The outgoing test signals are out-of-band of optical data signals traversing the ROADM. The ROADM block includes a test signal monitor configured to monitor for incoming test signals at each of the internal ports. The test signal monitor is configured to validate, based on a signature of an incoming test signal received at an internal port of the ROADM block, whether a valid connection exists between the internal port and an internal port of a second ROADM block.

Abstract: A server rack seismic restraint includes a rear restraint assembly configured to attach to a support surface that is operable to support a data center server rack. The rear restraint assembly includes at least one lateral member positioned to receive the server rack and restrain the server rack against lateral movement based on a force applied to the server rack; and an angled bracket positioned to receive the server rack and restrain the server rack against vertical movement based on the force applied to the server rack. The server rack seismic restraint also includes a front restraint assembly configured to attach to the support surface and engage the server rack to restrain the server rack against vertical movement based on the force applied to the server rack.

Abstract: Methods, systems, and media for protecting and verifying video files are provided.

Abstract: A transducer (140) having a mechanical impedance over an operative frequency range and having a desired power coupling (145) to a load over the operative frequency range comprises a piezoelectric device (141) having a frequency distribution of modes in the operative frequency range; and an overmould (143). The overmould (143) is arranged to surround at least part of the piezoelectric device (141); and the parameters of the overmould (143) are selected to provide a required impedance matching between the mechanical impedance of the transducer (140) and the mechanical impedance of the load. An alternative transducer comprises a mounting means for holding a discrete portion of at least a part of the periphery of the piezoelectric device wherein the parameters of the mounting means are selected to provide a required boundary condition for the periphery of the piezoelectric device whereby the desired power coupling between the transducer and the load is provided.

Abstract: Techniques for cooling a data center include circulating an airflow, to a warm air plenum of a first module, from rows of racks that support a heat-generating electronic devices; warming the airflow circulated through the racks; circulating the warmed airflow through a warm air inlet of the warm air plenum that is adjacent an open side of the racks and to a warmed air outlet adjacent a data center volume above the racks; circulating the airflow, with a fan positioned in a second module positioned in the data center volume above the racks, through at least one cooling module to cool the warmed airflow, and into a human-occupiable workspace of the data center adjacent the racks; and diverting the warmed airflow with an airflow partition mounted in the data center volume above the racks and adjusted to interrupt the warmed airflow between the warmed air plenum and the human-occupiable workspace.

Abstract: Methods, systems, and media for controlling network connections are provided. In some implementations, a method for controlling network connections is provided, the method comprising: determining, by a user device connection to an access point by a first network connection, that a second network connection is available; determining that the user device is in an idle state; while the user device is in the idle state, switching from the first network connection to the second network connection; monitoring network activity using the second network connection; switching back to the first network connection; generating a profile for the second network connection based at least in part on the monitored network activity; comparing the profile for the second network connection to a profile for the first network connection; selecting the second network connection based on the comparison; and in response to selecting the second network connection, switching to the second network connection.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for hotword detection on multiple devices are disclosed. In one aspect, a method includes the actions of receiving audio data that corresponds to an utterance. The actions further include determining that the utterance likely includes a particular, predefined hotword. The actions further include transmitting (i) data indicating that the computing device likely received the particular, predefined hotword, (ii) data identifying the computing device, and (iii) data identifying a group of nearby computing devices that includes the computing device. The actions further include receiving an instruction to commence speech recognition processing on the audio data. The actions further include in response to receiving the instruction to commence speech recognition processing on the audio data, processing at least a portion of the audio data using an automated speech recognizer on the computing device.

Abstract: A modular data center system includes an upper module including a roller assembly attached to a frame of the upper module, the roller assembly configured to engage at least a portion of a rail assembly attached to a structure of a data center building; and a lower module attachable to the upper module and configured to suspend above a floor of a human-occupiable workspace of the data center that supports a plurality of racks including a plurality of heat-generating electronic devices, the lower module including a power support assembly configured to provide electrical power to the plurality of heat-generating electronic devices.

Abstract: Methods, systems and apparatus for identifying illegitimate selections of content items. In some implementations, one or more servers can receive display data specifying a display state of a web page in a viewport. The web page includes a content item. Display instances are identified. A display instance is a display of at least a portion of the content item in the viewport. Selection instances of the content item are identified. A selection instance is a selection of the content item. The server(s) determines whether a selection of the content item occurred during a display of at least a portion of the content item in the viewport based on the display instances and the selection instances. A selection of the content is defined as an illegitimate selection if the selection did not occur during a display of at least a portion of the content item in the viewport.

Abstract: Antenna switch diversity circuitry can include four switches. A first switch can be connectable to a first transmitter, a first receiver, and a second receiver. A second switch can be connectable to a second transmitter, a third receiver, and a fourth receiver. A third switch can be directly connected to the first switch, the second switch, a first antenna, and a second antenna. A fourth switch directly can be connected to the first switch, the second switch, a third antenna, and a fourth antenna. The first and second switches can be configured to be controlled by a control signal in a manner so as to prevent a signal from the first transmitter and a signal from the second transmitter from being conveyed through a same switch, which can reduce a production of an intermodulation distortion signal by the signal from the first transmitter and the signal from the second transmitter.

Abstract: A method of providing cooled air to electronic equipment includes capturing heated air from a volume containing electronic equipment, cooling the heated air by more than fifteen degrees Celsius in an air-to-water heat exchanger, and supplying cooling water to the air-to-water heat exchanger at a temperature above a dew point temperature of the heated air.

Abstract: A method includes receiving an utterance of speech and segmenting the utterance of speech into a plurality of segments. For each segment of the utterance of speech, the method also includes extracting a speaker=discriminative embedding from the segment and predicting a probability distribution over possible speakers for the segment using a probabilistic generative model configured to receive the extracted speaker-discriminative embedding as a feature input. The probabilistic generative model trained on a corpus of training speech utterances each segmented into a plurality of training segments. Each training segment including a corresponding speaker-discriminative embedding and a corresponding speaker label. The method also includes assigning a speaker label to each segment of the utterance of speech based on the probability distribution over possible speakers for the corresponding segment.

Abstract: Selection and serving of content items may include receiving data indicative of a status of an energy source of a device with a request for a content item. A first received content item may be associated with a first energy consumption level and a second received content item may be associated with a second energy consumption level. The accessed content items are responsive to the request for a content item. The first energy consumption level may be higher than the second energy consumption level. The first content item or the second content item may be selected based, at least in part, on the received data indicative of the status of the energy source of the device, and data to display the selected content item may be provided to the device.

Abstract: A user captures images on a user computing device. The user signs in to an application, which transmits the user's images to an account management system, which recognizes objects within the images and assigns one or more object categories to the images and recognizes multiple images comprising objects in a common object category. After receiving user consent, the application groups the images on the user computing device according to object category. The user computing device captures an image of another object. The application transmits the image to the account management system, which detects objects within the image, identifies the object category, and saves the image to the corresponding object category group on the user computing device. After receiving user consent, the account management system finds information for each image in the object category group of images and transmits the information to the user computing device.

Abstract: Some implementations can include a computer-implemented method and/or system for parsing an electronic conversation for presentation at least partially in an alternative interface (e.g., a non-display interface) such as a voice interface or other non-display interface.

Abstract: A method of providing an ad extension includes selecting an advertisement for display. The method also includes selecting additional information related to the advertisement. The method also includes transmitting data representing the advertisement to a browser. The browser interacts with an expandable API to render an inline frame having an advertisement slot. The browser renders and displays the advertisement in the frame. The method also includes transmitting display data representing the additional information related to the advertisement to the browser. The browser receives an input to activate the ad extension. In response to the input, the browser interacts with the expandable API system to expand and render the frame. The browser renders, in the frame, the advertisement slot containing the advertisement. The browser also renders, in the frame, the additional information. The browser displays the expanded inline frame, such that the displayed frame covers a portion of the content.