Abstract: Various arrangements for performing ballistocardiography using a mobile device are presented. A radar integrated circuit of a mobile device may emit frequency-modulated continuous-wave (FMCW) radar. Reflected radio waves based on the FMCW radar being reflected off objects may be received and used to create a raw radar waterfall. The raw radar waterfall may be analyzed to create a ballistocardiography waveform. Data based on the ballistocardiography waveform may be output, such as to a machine-learning application installed on the mobile device.

Abstract: A direct speech-to-speech translation (S2ST) model includes an encoder configured to receive an input speech representation that to an utterance spoken by a source speaker in a first language and encode the input speech representation into a hidden feature representation. The S2ST model also includes an attention module configured to generate a context vector that attends to the hidden representation encoded by the encoder. The S2ST model also includes a decoder configured to receive the context vector generated by the attention module and predict a phoneme representation that corresponds to a translation of the utterance in a second different language. The S2ST model also includes a synthesizer configured to receive the context vector and the phoneme representation and generate a translated synthesized speech representation that corresponds to a translation of the utterance spoken in the different second language.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for training a deep neural network. One of the methods includes generating a plurality of feature vectors that each model a different portion of an audio waveform, generating a first posterior probability vector for a first feature vector using a first neural network, determining whether one of the scores in the first posterior probability vector satisfies a first threshold value, generating a second posterior probability vector for each subsequent feature vector using a second neural network, wherein the second neural network is trained to identify the same key words and key phrases and includes more inner layer nodes than the first neural network, and determining whether one of the scores in the second posterior probability vector satisfies a second threshold value.

Abstract: Implementations set forth herein relate to providing selectable autofill suggestions, which correspond to application actions that are at least partially fulfilled using server command data—prior to a user selecting a particular selectable autofill suggestion. Proactively fulfilling command data in this way mitigates latency between user selection of a suggestion and fulfillment of a particular action. Initially, a partial input can be processed to generate autofill suggestions, which can be communicated to a server device for further processing. The autofill suggestions can also be rendered for selection at a touch display interface, thereby allowing a user to select one of the autofill suggestions. As command fulfillment data is provided by the server, the command fulfillment data can be available to a corresponding application(s) in order that any corresponding actions can be at least partially fulfilled prior to user selection.

Abstract: Methods, systems, and media for sending a message about a new video to a group of related users are provided. In accordance with some embodiments, the method comprises: determining that a new video associated with a content creator has been uploaded to a video sharing service; identifying a group of candidate users to receive a message indicating that the new video is available; identifying a sub-group of users within the group of candidate users based on relationships of users within the group to each other; determining a time to send the message; transmitting the message to the sub-group of users at the determined time; determining that the message has been selected by a user device associated with a user of the sub-group of users; and causing the new video to be presented in a user interface including a communication interface associated with the sub-group of users on the user device.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for generating a computer chip placement. One of the methods includes obtaining netlist data for a computer chip; and generating a computer chip placement, comprising placing a respective macro node at each time step in a sequence comprising a plurality of time steps, the placing comprising, for each time step: generating an input representation for the time step; processing the input representation using a node placement neural network having a plurality of network parameters, wherein the node placement neural network is configured to process the input representation in accordance with current values of the network parameters to generate a score distribution over a plurality of positions on the surface of the computer chip; and assigning the macro node to be placed at the time step to a position from the plurality of positions using the score distribution.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for language models using domain-specific model components. In some implementations, context data for an utterance is obtained. A domain-specific model component is selected from among multiple domain-specific model components of a language model based on the non-linguistic context of the utterance. A score for a candidate transcription for the utterance is generated using the selected domain-specific model component and a baseline model component of the language model that is domain-independent. A transcription for the utterance is determined using the score the transcription is provided as output of an automated speech recognition system.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for distributing machine learning workloads, e.g., computations for training a neural network or computing an inference using a neural network, across multiple hardware accelerators.

Abstract: A highlight learning technique is provided to detect and identify highlights in sports videos. A set of event models are calculated from low-level frame information of the sports videos to identify recurring events within the videos. The event models are used to characterize videos by detecting events within the videos and using the detected events to generate an event vector. The event vector is used to train a classifier to identify the videos as highlight or non-highlight.

Abstract: A wireless accessory has one or more sensors positioned at an outer periphery of its housing, thereby forming a knob-like input. The sensors may be configured to detect a change in contact force or contact location of first and second fingertips at respective first and second opposite sides of the periphery. An input gesture is determined based on the detected change, and a command corresponding to the gesture may be executed.

Abstract: A device includes: a plurality of qubits arranged in a two-dimensional array and a plurality of readout resonators. Each readout resonator of a first readout resonator group is arranged to electromagnetically couple to a respective qubit of a first qubit group. Each readout resonator of a second readout resonator group is arranged to electromagnetically couple to a respective qubit of a second qubit group. A resonance frequency of each readout resonator of the first readout resonator group is within a first resonance frequency band, and a resonance frequency of each readout resonator of the second readout resonator group is within a second resonance frequency band that is different from the first resonance frequency band.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for obtaining, by a first sequence-training speech model, a first batch of training frames that represent speech features of first training utterances; obtaining, by the first sequence-training speech model, one or more first neural network parameters; determining, by the first sequence-training speech model, one or more optimized first neural network parameters based on (i) the first batch of training frames and (ii) the one or more first neural network parameters; obtaining, by a second sequence-training speech model, a second batch of training frames that represent speech features of second training utterances; obtaining one or more second neural network parameters; and determining, by the second sequence-training speech model, one or more optimized second neural network parameters based on (i) the second batch of training frames and (ii) the one or more second neural network parameters.

Abstract: Methods, systems, and media for delivering manifestless streaming media content are provided.

Abstract: Systems and methods are provided for a content-based security for computing devices. An example method includes identifying content rendered by a mobile application, the content being rendered during a session, generating feature vectors from the content and determining that the feature vectors do not match a classification model. The method also includes providing, in response to the determination that the feature vectors do not match the classification model, a challenge configured to authenticate a user of the mobile device. Another example method includes determining a computing device is located at a trusted location, capturing information from a session, the information coming from content rendered by a mobile application during the session, generating feature vectors for the session, and repeating this until a training criteria is met. The method also includes training a classification model using the feature vectors and authenticating a user of the device using the trained classification model.

Abstract: In a distributed system where a client's call to commit a transaction occurs outside the transaction's lock-hold interval, computation of timestamp information for the transaction is moved to a client library, while ensuring that no conflicting reads or writes are performed between a time of the computation and acquiring all locks for the transaction. The transaction is committed in phases, with each phase being initiated by the client library. Timestamp information is added to the locks to ensure that timestamps are generated during lock-hold intervals. An increased number of network messages is thereby overlapped with a commit wait period in which a write in a distributed database is delayed in time to ensure concurrency in the database.

Abstract: An electronic device measures noise variability of background noise present in a sampled audio signal, and determines whether the measured noise variability is higher than a high threshold value or lower than a low threshold value. If the noise variability is determined to be higher than the high threshold value, the device categorizes the background noise as having a high degree of variability. If the noise variability is determined to be lower than the low threshold value, the device categorizes the background noise as having a low degree of variability. The high and low threshold values are between a high boundary point and a low boundary point. The high boundary point is based on an analysis of files including noises that exhibit a high degree of variability, and the low boundary point is based on an analysis of files including noises that exhibit a low degree of variability.

Abstract: This technology relates to rasterizing and compositing vector graphics in parallel on a data-parallel computing device. For example, vector data of the vector graphics, may be loaded into local memory accessible by the one or more parallel processors. The vector data may include one or more paths comprised of one or more path segments of the vector graphics. The one or more parallel processors may rasterize the one or more path segments into respective rasters and assign each of the rasters into groups based on pixel coordinates of the respective rasters. Each group may have an associated key and the rasters within each group represent a portion of the same vector graphic. The rasters may be placed onto subpixels according to their respective pixel coordinates and rendered onto a display by the one or more parallel processors.

Abstract: Apparatus and methods related providing application execution information (AEI) are provided. A server can receive a request to provide a software package for a particular software application. The server can determine composite AEI (CAEI) for the particular software application. The CAEI can include a composite list of software with data about software methods of the particular software application executed by a computing device other than the server. The server can extract particular AEI related to the particular software application from the CAEI. The particular AEI can provide compiler hints for compiling at least one software method predicted to be executed by the particular software application. The server can generate the software package, where the software package can include the particular software application and the particular AEI. The server can provide the software package.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for optimizing integrated circuit architectures or compiler designs using an optimization engine. The optimization engine includes an auto-encoder and one or more regressors. Once trained, the optimization engine can encode initial, discrete input values of a set of input characteristics into a continuous domain and use continuous optimization techniques to identify final input values of the set of input characteristics that optimize one or more output characteristics.

Abstract: Replicas are selected in a large distributed network, and the roles for these replicas are identified. In one example, a leader is selected from among candidate computing dusters. To make this selection, an activity monitor predicts or monitors the workload of one or more clients. Different activities of the workload are given corresponding weights. The delay in performing requested activities, modified by these weights is found, and the candidate leader with the lowest weighted delay is selected as the leader.