Abstract: Techniques of generating recrawl policies for commercial offer pages include generating a multiple strategy approach using a number of different strategies. In some implementations, each strategy is an arm of a K-armed adversarial bandits algorithm with reinforcement learning. Moreover, in some implementations, the multiple strategy approach also uses a machine learning algorithm to estimate parameters such as a click rate, impression rate, and likelihood of price change, i.e., change rate, which was assumed known in the conventional approaches.

Abstract: Systems and methods of image searching include receiving content, receiving a request to select an image from content, selecting a plurality of items in the image, retrieving information about the selected item, and providing display data based on the retrieved information.

Abstract: A system for using mobile data to improve weather information is provided. The system includes a weather prediction station configured to receive stationary observation data provided by a plurality of stationary weather stations along with data from a plurality of input weather models and generate unified weather model estimates based on the stationary observation data, the input weather model data, and a processor. The processor is configured to aggregate mobile observation data provided by a plurality of non-stationary sensors and use the aggregated mobile observation data to adjust the weather model estimates.

Abstract: Game decisions are coordinated using a semantic natural language processing (NLP) machine learning (ML) algorithm, which is stored in a memory in some cases. In response to a game event, a processor records a text string that represents the game event in a text log that includes a sequence of text strings that represent game events that have transpired during a portion of the game. The processor also generates, using the semantic NLP ML algorithm, scores for labeled actions or content based on the text log and a curve that represents a target player experience as a function of progress through the game. The processor further serves one or more of the labeled actions or content that is selected based on the scores. The labeled actions or content are served to a display associated with the processor.

Abstract: Systems and methods according to the present disclosure can employ a computer-implemented method for inference using a machine-learned model. The method can be implemented by a computing system having one or more computing devices. The method can include obtaining data descriptive of a neural network including one or more network units and one or more gating paths, wherein each of the gating path(s) includes one or more gating units. The method can include obtaining data descriptive of one or more input features. The method can include determining one or more network unit outputs from the network unit(s) based at least in part on the input feature(s). The method can include determining one or more gating values from the gating path(s). The method can include determining one or more gated network unit outputs based at least in part on a combination of the network unit output(s) and the gating value(s).

Abstract: Implementations are directed to training a machine learning model that, once trained, is used in performance of robotic grasping and/or other manipulation task(s) by a robot. The model can be trained using simulated training examples that are based on simulated data that is based on simulated robot(s) attempting simulated manipulations of various simulated objects. Portion(s) of the model can also be trained based on real training examples that are based on data from real-world physical robots attempting manipulations of various objects. The simulated training examples can be utilized to train the model to predict an output that can be utilized in a particular task—and the real training examples used to adapt at least a portion of the model to the real-world domain can be tailored to a distinct task. In some implementations, domain-adversarial similarity losses are determined during training, and utilized to regularize at least portion(s) of the model.

Abstract: The present disclosure provides systems and methods that use and/or generate image files according to a novel microvideo image format. For example, a microvideo can be a file that contains both a still image and a brief video. The microvideo can include multiple tracks, such as, for example, a separate video track, audio track, and/or one or more metadata tracks. As one example track, the microvideo can include a motion data track that stores motion data that can be used (e.g., at file runtime) to stabilize the video frames. A microvideo generation system included in an image capture device can determine a trimming of the video on-the-fly as the image capture device captures the microvideo.

Abstract: The technology involves messaging applications that support message threads that can involve grouped messages pertaining a conversation, post, or topic. A user interface can employ a communication pane or other message area to present all threads in a particular room, not just the threads that the user is currently following. Following can occur when a user elects to be notified of new replies by clicking a “follow” indicia, replying to or starting a thread, or when someone else specifically mentions the user. The UI may present different notifications about new replies or other details based on a type of display device the person is currently using. For threads that include specific references to the user or are not currently being followed, graphical indicia can be presented as a tailored alert. Different filter options in the user interface enable the user to see threads they are following or mentioned in.

Abstract: Systems and methods provide an application programming interface to offer action suggestions to third-party applications using context data associated with the third-party. An example method includes receiving content information and context information from a source mobile application, the content information representing information to be displayed on a mobile device as part of a source mobile application administered by a third party, the context information being information specific to the third party and unavailable to a screen scraper. The method also includes predicting an action based on the content information and the context information, the action representing a deep link for a target mobile application. The method further includes providing the action to the source mobile application with a title and a thumbnail, the source mobile application using the title and thumbnail to display a selectable control that, when selected, causes the mobile device to initiate the action.

Abstract: The present disclosure is directed to selective gesture recognition for handheld device gestures. An example method includes receiving, by a handheld interactive object, movement information descriptive of a gesture performed with the handheld interactive object. The method includes selecting a local and/or remote machine-learned model for processing the movement information. The movement information can be processed to identify a gesture action corresponding to the movement information. The local and/or remote machine-learned model can be selected based on user input data and/or a complexity of the movement information. In response to selecting the local machine-learned model, the method includes processing the movement information according to the local machine-learned model and communicating a message to a remote device based on the result.

Abstract: An example method includes receiving, by a computational assistant executing at one or more processors, a representation of an utterance spoken at a computing device; identifying, based on the utterance, a task to be performed by the computational assistant; responsive to determining, by the computational assistant, that complete performance of the task will take more than a threshold amount of time, outputting, for playback by one or more speakers operably connected to the computing device, synthesized voice data that informs a user of the computing device that complete performance of the task will not be immediate; and performing, by the computational assistant, the task.

Abstract: Granting access to a communication channel for a user device by a base station includes receiving, from a user device, a request to access the communication channel, the request including a random access preamble and a payload, attempting to decode the payload, selecting a format of a response message in accordance with whether the payload was decoded successfully, and transmitting, to the user device, the response message, including providing an indication of the format of the response message to the user device.

Abstract: Generating and/or recommending command bundles for a user of an automated assistant. A command bundle comprises a plurality of discrete actions that can be performed by an automated assistant. One or more of the actions of a command bundle can cause transmission of a corresponding command and/or other data to one or more devices and/or agents that are distinct from devices and/or agents to which data is transmitted based on other action(s) of the bundle. Implementations determine command bundles that are likely relevant to a user, and present those command bundles as suggestions to the user. In some of those implementations, a machine learning model is utilized to generate a user action embedding for the user, and a command bundle embedding for each of a plurality of command bundles. Command bundle(s) can be selected for suggestion based on comparison of the user action embedding and the command bundle embeddings.

Abstract: The technology manages message threads. A computing system is configured to detect at least one question in a message thread and to determine if the message thread includes at least one answer responding to the at least one question. From this, the system generates a summary of the message thread based on the detection of the at least one question and the determination of the at least one answer, and outputs for display the summary of the message thread to a recipient of the message thread. The output may involve visually indicating the detected at least one question and/or visually indicating the at least one answer responding to the at least one question. The detected at least one question and the at least one answer may be visually indicated in a manner distinct from remaining content of the message thread.

Abstract: Training and/or using a machine learning model for performing robotic tasks is disclosed herein. In many implementations, an environment-conditioned action sequence prediction model is used to determine a set of actions as well as a corresponding particular order for the actions for the robot to perform to complete the task. In many implementations, each action in the set of actions has a corresponding action network used to control the robot in performing the action.

Abstract: A user device (UE) for scheduling an uplink transmission assignment with a base station that communicates with the user device via a shared carrier, receives, from the base station, a configuration that indicates (i) a first channel access procedure for the user device to perform prior to transmitting an uplink transmission and (ii) at least one occasion at which the user device is to transmit the uplink transmission (1602); receives, from the base station via the shared carrier, a signal indicating at least a portion of a transmission time period during which the shared carrier is available to the base station (1604); and performs the first channel access procedure or a second channel access procedure before transmitting the uplink transmission based at least in part on whether the occasion is within the transmission time period (1606).

Abstract: Systems and methods for validation of modeling and simulation systems that provide for the virtual try-on of wearable devices, such as glasses, by a user, and for the virtual fitting of selected wearable devices for the user. Wearable fit measurements, display fit measurements, ophthalmic fit measurements and other such measurements associated with the fit and function of the wearable device may be detected from image data capturing the wearable device worn by the user. The detected measurements may be compared to corresponding measurements detected in a virtual simulation of the wearable device worn by the user. The comparison may provide for validation and increased accuracy/realism of the modeling and simulation systems.