Abstract: Methods, systems, and apparatus, for performing low-power vision sensing. One computing device includes a vision sensor configured to generate vision sensor data and an ambient computing system configured to repeatedly process the vision sensor data generated by the vision sensor according to a low-power detection process. If a detection is indicated by the low-power detection process, the ambient computing system wakes one or more other components of the computing device to perform a high-power detection process using the vision sensor data.

Abstract: Systems and methods for generating geographic data for map elements based on surfel data and motion data associated with a geographic region are disclosed. A computing system can obtain motion data indicative of movement of at least one object in a geographic region represented by a map. The computing system can additionally obtain surfel data indicative of one or more surface elements associated with a surface of a structure in the geographic region represented by the map. The computing system can identify an entrance of the structure based at least in part on a correlation of the surfel data and the motion data. The computing system can generate geographic data indicative of a geographic location of the entrance of the structure.

Abstract: This document describes techniques and systems for radar-based gesture-recognition with context-sensitive gating and other context-sensitive controls. Sensor data from a proximity sensor and/or a movement sensor produces a context of a user equipment. The techniques and systems enable the user equipment to recognize contexts when a radar system can be unreliable and should not be used for gesture-recognition, enabling the user equipment to automatically disable or “gate” the output from the radar system according to context. The user equipment prevents the radar system from transitioning to a high-power state to perform gesture-recognition in contexts where radar data detected by the radar system is likely due to unintentional input. By so doing, the techniques conserve power, improve accuracy, or reduce latency relative to many common techniques and systems for radar-based gesture-recognition.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for indicating callers for incoming voice calls. The methods, systems, and apparatus include actions receiving an incoming voice call, determining a calling number and a called number from the incoming voice call, identifying a user account that corresponds to the called number, determining a contact name for the calling number based on contact entries for the user account, and providing the contact name for output.

Abstract: This document describes techniques and apparatuses for optimizing a cellular network using machine learning. In particular, a network-optimization controller uses machine learning to determine an optimized network-configuration parameter that affects a performance metric of the cellular network. To make this determination, the network-optimization controller requests and analyzes gradients determined by one or more user equipments, one or more base stations, or combinations thereof. By using machine learning, the network-optimization controller identifies different optimized network-configuration parameters associated with different local optima or global optima of an optimization function, and selects a particular optimized network-configuration parameter that is appropriate for a given environment. In this manner, the network-optimization controller dynamically optimizes the cellular network to account for both short-term and long-term environmental changes.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for performing a machine learning task on a network input to generate a network output. In one aspect, one of the systems includes a neural network configured to perform the machine learning task, the neural network including one or more switch layers.

Abstract: Aspects of the disclosed technology include a method including receiving, from a user device, an identification of content; receiving, by a computing device, the identified content; accessing search engine processing logic; processing the received content using the subset of search engine processing logic, without indexing the received content to be accessed for responding to search queries from the search engine; generating a representation of a predicted search result of the received content based on the processing; and transmitting, to the user device, the representation of the predicted search result.

Abstract: A method performed by a server can include sending a first webpage to a first computing device, the first computing device including a camera, the first webpage including an image-capturing function and including an instruction for a user to obtain a second webpage via a second computing device, the second webpage including a calibration image file; receiving, from the first computing device, multiple captured images that were captured by the camera, the multiple calibration images including instances of a calibration image presented by a display included in the second computing device, the calibration image being a representation of the calibration image file; and based on the multiple captured images, calibrating the camera.

Abstract: Techniques are described herein for iterative code generation using neural language models. In various implementations, an original source code snippet in a first programming language may be processed using a translation machine learning model to generate a first translation of the original source code snippet in a second programming language. The first translation of the original source code snippet may be evaluated to identify error(s) in the first translation. Based on the error(s), respective mask(s) may be inserted to generate a masked first translation of the original source code snippet in the second programming language. The masked first translation of the original source code snippet may be processed using the translation machine learning model to generate a second translation of the original source code snippet in the second language. The second translation may include infill(s) of corrected source code in place of one or more of the masks.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for detecting hotwords using a server. One of the methods includes receiving an audio signal encoding one or more utterances including a first utterance; determining whether at least a portion of the first utterance satisfies a first threshold of being at least a portion of a key phrase; in response to determining that at least the portion of the first utterance satisfies the first threshold of being at least a portion of a key phrase, sending the audio signal to a server system that determines whether the first utterance satisfies a second threshold of being the key phrase, the second threshold being more restrictive than the first threshold; and receiving tagged text data representing the one or more utterances encoded in the audio signal when the server system determines that the first utterance satisfies the second threshold.

Abstract: A first computing device captures audio data via an audio capture device. The audio data comprises probe signals from second computing devices located within the same area. The first computing device and the second computing devices are connected to a teleconference session. Based on the audio data, the first computing device generates a first encoding of the probe signals received. The first computing device receives a second encoding of probe signals received at a second computing device. The first computing device makes a determination that a degree of similarity between the first encoding and the second encoding is greater than or equal to a threshold degree of similarity. The first computing device generates co-location information identifying the second computing device as a co-located device that is located within the same area as the first computing device.

Abstract: Methods, systems, and storage media for classifying content across media formats based on weak supervision and cross-modal training are disclosed. The system can maintain a first feature classifier and a second feature classifier that classifies features of content having a first and second media format, respectively. The system can extract a feature space from a content item using the first feature classifier and the second feature classifier. The system can apply a set of content rules to the feature space to determine content metrics. The system can correlate a set of known labelled data to the feature space to construct determinative training data. The system can train a discrimination model using the content item and the determinative training data. The system can classify content using the discrimination model to assign a content policy to the second content item.

Abstract: Techniques are described herein for translating source code using sparse-self attention. In various implementations, a source code snippet in a first programming language may be processed to obtain graph(s) representing snippet tokens, and relationships therebetween. Based on the graph(s), a subset of snippet token pairs may be identified from a superset of all possible token pairs in the source code snippet. Each token pair of the subset may include snippet tokens that are represented by nodes connected by one or more edges of the one or more graphs. A self-attention network of a translation machine learning model may be adapted to sparsely attend across the identified subset of token pairs. The source code snippet may then be processed based on the adapted translation machine learning model to generate a translation of the source code snippet in the second programming language.

Abstract: A computer-implemented method of training a streaming speech recognition model that includes receiving, as input to the streaming speech recognition model, a sequence of acoustic frames. The streaming speech recognition model is configured to learn an alignment probability between the sequence of acoustic frames and an output sequence of vocabulary tokens. The vocabulary tokens include a plurality of label tokens and a blank token. At each output step, the method includes determining a first probability of emitting one of the label tokens and determining a second probability of emitting the blank token. The method also includes generating the alignment probability at a sequence level based on the first probability and the second probability. The method also includes applying a tuning parameter to the alignment probability at the sequence level to maximize the first probability of emitting one of the label tokens.

Abstract: Examples relate to implementations of a neural light transport. A computing system may obtain data indicative of a plurality of UV texture maps and a geometry of an object. Each UV texture map depicts the object from a perspective of a plurality of perspectives. The computing system may train a neural network to learn a light transport function using the data. The light transport function may be a continuous function that specifies how light interacts with the object when the object is viewed from the plurality of perspectives. The computing system may generate an output UV texture map that depicts the object from a synthesized perspective based on an application of the light transport function by the trained neural network.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for retrieving and using contextual data from previous conversation sessions in conversational searches. In one aspect, a method includes receiving a first query for a first user session, determining that the first query refers to one or more tags in a first repository, the first repository associating respective identifiers to respective tags, each identifier representing a corresponding user session, determining one or more particular identifiers associated with the one or more tags in the first repository, retrieving particular contextual data associated with the determined particular identifiers in a second repository, the second repository associating respective identifiers to respective contextual data associated with corresponding user sessions represented by the respective identifiers, and performing an action responsive to the first query based on the retrieved particular contextual data.

Abstract: Syntax elements are written to a bitstream to designate bit depth precision for palette mode coding of video blocks. During encoding, a bit depth to use for palette mode coding a current block may be based on an input video signal including the current block or based on some change in bit depth precision. A prediction residual for the current block is encoded to a bitstream along with syntax elements indicative of the bit depth used for the palette mode coding of the current block. In particular, the syntax elements include a first element indicating the palette mode coding bit depth used and a second element indicating whether to apply a bit offset to the palette mode coding bit depth. During decoding, values of the syntax elements are read from the bitstream and used to determine a bit depth for palette mode coding the encoded block.

Abstract: A method includes receiving training data including transcribed speech utterances spoken in a general domain, modified speech utterances in a target domain, and unspoken textual utterances corresponding to the transcriptions of the modified speech utterances in the target domain. The modified speech utterances include utterances spoken in the target domain that have been modified to obfuscate one or more classes of sensitive information recited in the utterances. The method also includes generating a corresponding alignment output for each unspoken textual utterance of the received training data using an alignment model. The method also includes training a speech recognition model on the alignment outputs generated for the corresponding to the unspoken textual utterances, the un-transcribed speech utterances, and the transcribed speech utterances to teach the speech recognition model to learn to recognize speech in the target domain and phrases within the one or more classes of sensitive information.