Abstract: A method may include (i) obtaining first data records structured in accordance with a first schema, (ii) determining, for the first schema, one or more first schema property values for each schema property in a set of pre-defined schema properties, (iii) determining, for a second schema, one of more second schema property values for each schema property in the set of pre-defined schema properties, (iv) providing, to a schema matching engine, first and second schema property values, where the schema matching engine contains schema mapping techniques and rules, where each rule suggests a schema mapping technique based on schema properties from the set of pre-defined schema properties, (v) applying the rules to select a schema mapping technique, (vi) transforming the first data records in accordance with the selected schema mapping technique, and (vii) providing the transformed first data records in a data structure in accordance with the second schema.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for adjusting physical location usage for a plurality of particular locations. Methods can include obtaining a three-dimensional (3D) representation of the given geographic area, wherein the 3D representation depicts a view of the given geographic area from a specified viewing perspective. For the given geographic area, traffic data is obtained indicating different traffic volumes during different time periods and one or more traffic characteristics. The 3D representation is segmented into a plurality of particular locations. For each particular location among the plurality of particular locations and based on the traffic data, a viewability score is determined that indicates an aggregate amount of time that the particular location is viewable by traffic passing the different locations. Physical location usage is then adjusted based on the viewability scores for the plurality of particular locations.

Abstract: A method for load balancing application requests across a multi-cluster container orchestration system includes receiving a load-balancing configuration for a multi-cluster service managing access to a set of destination clusters hosting a software application deployed by a user. The multi-cluster service uses the load-balancing configuration to load balance application level traffic across the set of destination clusters. Each destination cluster includes at least one container executing the software application and a respective geographical region. The method also includes receiving an application level request directed toward the software application hosted across the destination clusters. The application level request is received from a client and includes a host name and a geographical location associated with the client. The method also includes routing the request to one of the clusters based on the geographical location of the request and the respective geographical locations of the clusters.

Abstract: This disclosure describes systems and techniques for synchronizing cameras and tagging images for face authentication. For face authentication by a facial recognition model, a dual infrared camera may generate an image stream by alternating between capturing a “flood image” and a “dot image” and tagging each image with metadata that indicates whether the image is a flood or a dot image. Accurately tagging images can be difficult due to dropped frames and errors in metadata tags. The disclosed systems and techniques provide for the improved synchronization of cameras and tagging of images to promote accurate facial recognition.

Abstract: Methods, systems, and media for presenting recommended media content items based on complementary media content item pairs are presented.

Abstract: A method for training hotword detection includes receiving a training input audio sequence including a sequence of input frames that define a hotword that initiates a wake-up process on a device. The method also includes feeding the training input audio sequence into an encoder and a decoder of a memorized neural network. Each of the encoder and the decoder of the memorized neural network include sequentially-stacked single value decomposition filter (SVDF) layers. The method further includes generating a logit at each of the encoder and the decoder based on the training input audio sequence. For each of the encoder and the decoder, the method includes smoothing each respective logit generated from the training input audio sequence, determining a max pooling loss from a probability distribution based on each respective logit, and optimizing the encoder and the decoder based on all max pooling losses associated with the training input audio sequence.

Abstract: A method includes receiving a plurality of unlabeled audio samples corresponding to spoken utterances not paired with corresponding transcriptions. At a target branch of a contrastive Siamese network, the method also includes generating a sequence of encoder outputs for the plurality of unlabeled audio samples and modifying time characteristics of the encoder outputs to generate a sequence of target branch outputs. At an augmentation branch of a contrastive Siamese network, the method also includes performing augmentation on the unlabeled audio samples, generating a sequence of augmented encoder outputs for the augmented unlabeled audio samples, and generating predictions of the sequence of target branch outputs generated at the target branch. The method also includes determining an unsupervised loss term based on target branch outputs and predictions of the sequence of target branch outputs. The method also includes updating parameters of the audio encoder based on the unsupervised loss term.

Abstract: This application is directed to a doorbell camera for illuminating and capturing scenes. The doorbell camera includes at least a subset of processors for operating a camera module, an image sensor having a field of view of a scene and configured to capture video of a portion of the scene, one or more infrared (IR) illuminators for providing illumination, a waterproof button assembly, and a microphone and a speaker for enabling a real-time conversation between a visitor located at the doorbell camera and a user of a remote client device. The waterproof button assembly is configured to receive a user press on a button top, block water from entering the electronic device, and display a visual pattern uniformly at a peripheral region of the button assembly using LEDs and light guide component that are disposed under the button top.

Abstract: Techniques and apparatuses are described that implement control signaling for monostatic radar sensing. In particular, a base station uses control signaling to configure a user equipment for monostatic radar sensing and control when monostatic radar sensing is performed by the user equipment. With control signaling, the base station can enable monostatic radar sensing to occur using similar frequency resources used for wireless communication, which enables efficient use of a frequency spectrum. The base station can also use control signaling to reduce interference observed by other user equipment as the user equipment performs monostatic radar sensing. By performing monostatic radar sensing, the user equipment compiles explicit information about objects within an operating environment and shares this information with the base station. The base station uses this information to improve wireless communication performance.

Abstract: A method includes receiving, at a first processor of a user device, streaming multi-channel audio captured by an array of microphones, each channel including respective audio features. For each channel, the method also includes processing, by the first processor, using a first stage hotword detector, the respective audio features to determine whether a hotword is detected. When the first stage hotword detector detects the hotword, the method also includes the first processor providing chomped raw audio data to a second processor that processes, using a first noise cleaning algorithm, the chomped raw audio data to generate a clean monophonic audio chomp. The method also includes processing, by the second processor using a second stage hotword detector, the clean monophonic audio chomp to detect the hotword.

Abstract: The subject matter described herein provides systems and techniques for controlling the production of eNoise in an audio playback system. The eNoise may be an audible noise produced when a current, such as a noise/ripple current, flows through a battery of the audio playback system. Such eNoise may be reduced by limiting the current, such as noise/ripple current, flowing through the battery. In some examples, the noise/ripple current may be diverted to the main board of the audio playback system by adding a source of current. A circuit to produce such a current may include transistor(s), capacitor(s), and/or resistor(s). Using the current source and/or such a circuit may also divert the noise/ripple current away from the battery, thereby reducing the eNoise produced.

Abstract: Aspects of the disclosure provide for natively executing row-store expression data structures on column-store databases without rewriting. A database management system (DBMS) configured as described herein can maintain a mapping of row-store results to addresses of where corresponding column data is stored. When executing operators, such as logical operators, comparison operators, and/or function operators of a received query expression, the DBMS can operate on the column data, rather than the individual rows. The DBMS can store the results generated by executing the column operators, for example on a stack, and record the row-store addresses to which the stored results correspond. The DBMS responds with a number of rows corresponding to the processed column data.

Abstract: Methods, systems, and media for providing dynamic media sessions with video stream transfer features are provided. In some embodiments, the methods include: receiving an indication that presentation of the video content item is to be transferred from the first display device to a different display device; in response to receiving the indication, determining whether the first display device supports transferring the presentation of the video content item and determining whether the different display device is capable of presenting the video content item; and transmitting an indication that presentation of the video content item is to be transferred from the first display device to the different display device, wherein the video content item is played back using the different display device from a playback position that is based on the playback state from the first display device.

Abstract: Methods, systems, and apparatus, including computer-readable storage media, for course assignment by a multi-learning management system. The system can receive data from a variety of individual learning management systems offering different courses. The system can use feedback data of a user base for the system to cluster courses by predicted difficulty, and generate, from the clusters, a sequence of courses for a target user. The sequence of courses can include at least one course from each cluster, with courses from a first cluster with a lower overall difficulty measure preceding courses in a second cluster with a higher overall difficulty measure in the sequence, wherein the starting cluster can be calculated according to the estimated level of the target user.

Abstract: Throughput is preserved in a distributed system while maintaining concurrency by pushing a commit wait period to client commit paths and to future readers. As opposed to servers performing commit waits, the servers assign timestamps, which are used to ensure that causality is preserved. When a server executes a transaction that writes data to a distributed database, the server acquires a user-level lock, and assigns the transaction a timestamp equal to a current time plus an interval corresponding to bounds of uncertainty of clocks in the distributed system. After assigning the timestamp, the server releases the user-level lock. Any client devices, before performing a read of the written data, must wait until the assigned timestamp is in the past.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for determining and recommending the types of digital components that content providers can generate and provide for distribution to client devices. In one aspect, a method can determine whether a content provider has not previously provided a first digital component of a first media type. A first set of user interaction data can be obtained and input into a machine learning model. The model can can output result data for expected affirmative user actions related to the first digital component of the first media type. Based on the result data, a recommendation specifying whether the content provider should provide the first digital component of the first media type can be generated and provided to the content provider.

Abstract: Methods, systems, and media for identifying abusive user accounts based on playlists are provided. In accordance with some embodiments of the disclosed subject matter, a method for identifying abusive content is provided, the method comprising: determining at least one feature associated with a playlist created by a user-generated channel; calculating a playlist score associated with the playlist based on a playlist classifier, wherein the playlist classifier comprises a function that maps the at least one feature to the playlist score; calculating a channel score associated with the user-generated channel based at least on the calculated playlist score; determining that one or more content items associated with the user-generated channel is to be demoted based on the calculated channel score, wherein the one or more content items comprises the playlist; and causing the one or more content items to be demoted.

Abstract: A Bluetooth audio streaming passthrough, and a method for transmitting an audio stream over a Bluetooth communication link and disclosed. The method is carried out by a source device. An audio codec configuration and a sampling rate of the audio stream are received, the audio stream coming from a streaming application running on the source device. A Bluetooth audio stream between the source device and a sink device is configured using the received audio codec configuration and sampling rate. The first audio stream is received from the streaming application. The received first audio stream is packetized into the Bluetooth audio stream. The Bluetooth audio stream is transmitted to the sink device over the Bluetooth communication link, including the packetized received audio stream.

Abstract: A data processing device incorporates a plurality of chiplets having working elements such as processing and memory elements. At least one of the working elements is operative to generate messages directed to working elements of the same chiplet or another one of the chiplets. Each message includes a global address. An evaluation circuit determines whether the global address of a message is within a range of global addresses assigned to the chiplet. If so, the message passes to a translation circuit which translates the message to a local address for routing to a working element of the chiplet. If not, the message is dispatched to one or more other chiplets.