Abstract: Embodiments maintain a data pool that includes heterogeneous data sets, and receiving a first data batch of a data set from a data source into the data pool. Embodiments determine a current state of the data set based on a data set state diagram including a plurality of data set states, and identify a condition of the first data batch. Embodiments further set a data batch state for the first data batch, based on a data batch state diagram, and update the data batch state of a prior data batch received before the first data batch, based on the condition of the first data batch. Embodiments additionally transition the data set state diagram, based on the condition of the first data batch, to an updated data set state. Embodiments maintain a data state repository storing the data set state for each of the plurality of heterogeneous data sets.

Abstract: A face-processing system is described for producing a target image based on a source image and driving information. The source image includes data depicting at least a face of a source subject having a source identity, a source pose, and a source expression. The driving information specifies one or more driving characteristics. The target image combines characteristics of the source image and the driving information. According to illustrative implementations, the face-processing system produces the target image by using plural warping subcomponents that operate at plural respective levels of a neural network and at increasing respective resolutions. Each warping subcomponent operates, in part, based on geometric displacement field (GDF) information that describes differences between a source mesh derived from the source image and a driving mesh derived from the driving information.

Abstract: The techniques disclosed herein enable systems to reduce the time required to terminate a set of microservices for an application while ensuring high availability and preventing request failures. This is accomplished through a termination manager which retrieves request queues for the microservices to analyze outstanding requests that require processing prior to termination. Based on the outstanding requests, the termination manager constructs call graphs for each request. The call graphs capture the operational flow of the associated request by defining a sequence of microservices whose functionality is invoked by the request. From an initial analysis, the termination manager can determine that some of the microservices do not appear in the call graphs, indicating that the microservices are not needed to process the outstanding requests. Accordingly, the unneeded microservices are terminated.

Abstract: Methods, systems, and computer programs are presented for selecting notifications based on an affinity score between a content generator and a viewer of the content. One method includes capturing interactions of content generators with notifications, received by the content generators, associated with viewer responses to creator-generated content items. The method further includes training a machine-learning model based on the interactions, and detecting a first set of notifications, for a first content generator, associated with interactions of a set of viewers to first-content generator content. The ML model calculates an affinity score between the first content generator and each viewer, and the set of first notifications are ranked based on the affinity scores of the first content generator and the viewer associated with each notification.

Abstract: Disclosed in some examples are methods, systems, devices, and machine-readable mediums which provide for sidebar communication threads forked from, or related to, a principal thread. Messages in the sidebar communication thread may include a history of the principal thread, including one or more messages from the principal thread, and may include a proposed principal thread message that is the subject of the sidebar thread discussion. The sidebar thread may also include sidebar thread messages that carries the conversation of the sidebar thread participants. Once a termination condition is reached for the sidebar thread, the sidebar thread terminates and either the proposed principal thread message (as potentially modified by participants of the sidebar thread) becomes an accepted principal thread message and it is posted to the principal thread as if it was sent by the sidebar initiator or no message is posted (e.g., the proposed principal thread message is rejected).

Abstract: An integrated fastener includes a threaded portion, a fastener head, and one or more spring fingers integrated into the fastener head, wherein the spring fingers are configured to apply a compressive load when the fastener is inserted and secured to a receiving assembly.

Abstract: Examples are disclosed that relate to generating expressive avatars using multi-modal three-dimensional face modeling and tracking. One example includes a computer system comprising a processor coupled to a storage system that stores instructions. Upon execution by the processor, the instructions cause the processor to receive initialization data describing an initial state of a facial model. The instructions further cause the processor to receive a plurality of multi-modal data signals. The instructions further cause the processor to perform a fitting process using the initialization data and the plurality of multi-modal data signals. The instructions further cause the processor to determine a set of parameters based on the fitting process, wherein the determined set of parameters describes an updated state of the facial model.

Abstract: This document relates to training and employing a quality estimation model. One example includes a method or technique that can be performed on a computing device. The method or technique can include providing degraded audio signals to one or more packet loss concealment models, and obtaining enhanced audio signals output by the one or more packet loss concealment models. The method or technique can also include obtaining quality labels for the enhanced audio signals and training a quality estimation model to estimate audio signal quality based at least on the enhanced audio signals and the quality labels.

Abstract: Examples are disclosed that relate to applying haptic output to a touch-sensitive input device. One example provides a touch-sensitive input device comprising a body, a haptic feedback mechanism within the body, a sensor subsystem, a logic processor, and a memory. The memory stores instructions executable by the processor to receive from the sensor subsystem sensor data indicating locations along the body of a plurality of contact points between a user hand and the body, based at least in part on the sensor data, determine a touch profile of the user hand applied to the body, based at least in part on the touch profile of the user hand, determine a selected haptic output to be applied to the body, and cause a drive signal to be transmitted to the haptic feedback mechanism to apply the selected haptic output to the body.

Abstract: Systems and methods for generating a shared collaborative channel for collaboration are provided. In particular, a computing device may receive a request, from an originating member of an organization, to create the shared collaborative channel, the request including an invitee to be added to the shared collaborative channel. In response to receipt of the request, the computing device may provision a substrate group by creating a container associate with the shared collaborative channel including a substrate database associated with the shared collaborative channel, generate an invitation including a custom link to the shared collaborative channel for the invitee, and determine whether the invitee belongs to an originating collaboration team associated with the originating member based on the substrate database. If the invitee belongs to the originating collaboration team, the computing device may further update the substrate group to add the invitee as a new member of the shared collaborative channel.

Abstract: A computing system includes a logic subsystem and a storage subsystem holding instructions executable by the logic subsystem to implement a transformer-based text encoder. The transformer-based text encoder includes a plurality of transformer blocks previously-trained to apply encoding operations to computer-readable text representations of input text strings, the computer-readable text representations including computer-readable question representations of input text questions, and computer-readable passage representations of input text passages. The plurality of transformer blocks include a shared transformer block trained for both the computer-readable question representations and the computer-readable passage representations and a specialized transformer block including two or more input-specific subnetworks, and a routing function to select an input-specific subnetwork of the two or more input-specific subnetworks for each of the computer-readable text representations.

Abstract: The description relates to accelerator architectures for deep learning models. One example can obtain a deep learning training script associated with a deep learning model and extract an operator graph from the training script. The example can split the operator graph into first and second portions of a heterogeneous pipeline and tune a first accelerator core for the first portion of the heterogeneous pipeline and a second accelerator core for the second portion of the heterogeneous pipeline. The example can also generate a hardware architecture that includes the first accelerator core and the second accelerator core arranged to collectively accomplish the deep learning model.

Abstract: Indirect time-of-flight camera systems for operating in multiple optical channels using active modulated light and accompanying methods of operation are provided. In one aspect, the indirect time-of-flight camera system includes first and second modulatable laser sources outputting light of different wavelengths for illuminating a target environment. The camera system further includes a wavelength-selective reflective element designed to reflect the light of a first wavelength and to transmit the light of a second wavelength. The camera system further includes a controller comprising instructions executable to control the camera system to, in a first time period, activate the first modulatable laser source and deactivate the second modulatable laser source, and in a second time period, deactivate the first modulatable laser source and activate the second modulatable laser source.

Abstract: Examples are disclosed that relate to fairly ordering financial market trades received from different market participant computers via a cloud computing network. In one example, a plurality of trades generated by a plurality of market participant computers are received. The trades are generated based at least on a financial market data point received by the plurality of market participant computers. Each trade is tagged with a delivery clock time stamp that tracks time in relation to financial market events that occur at a corresponding market participant computer. The trades are ordered based on the delivery clock time stamps and sent to a central exchange server computer. The central exchange server computer processes the trades.

Abstract: A general methodology is presented for optimizing a value at risk (VaR) associated with an allocation of objects (i.e., a strategy) having variable performance and loss characteristics. For purposes of illustration, investment strategies prescribing a portfolio of items from a set of candidates with unknown and generally correlated joint losses are discussed. The framework is based on approximating the VaR using nonparametric estimates of the portfolio loss density and, using mathematical insights, an efficient approach to computing the VaR gradient with respect to the strategy. The approach also allows inclusion of constraints on the strategy (e.g. a maximum fraction per item) and allows the VaR optimization problem to be solved using optimization techniques such as sequential quadratic programming.

Abstract: A foldable computing device comprises a first frame rotatably coupled to a second frame. The second frame comprises a push-to-open mechanism comprising an actuator and a power switch located for actuation by the actuator. A detection mechanism detects a displaced position of the actuator that corresponds to releasing the foldable computing device from a closed configuration. Actuation of the power switch is detected and used with detection of the displaced position of the actuator to control an operating state of the computing device.

Abstract: The description relates to resource aware object detection for encoded video streams that can identify frames of the video stream that include an object of interest, such as a human, without decoding the frames.

Abstract: Methods and systems cause display of email messages of a user on a screen of a computing system based on scores associated with the email messages. An email ranking system may have assigned the scores to the email messages. The scores are based on actions that other recipients of the email messages have taken with respect to the email messages. In calculating the scores, the actions of the other recipients may receive different weights based on how closely connected a recipient is to the user and a type of connection the recipient has to the user. A network graph may indicate how closely connected the recipient is to the user and the type of connection the recipient has to the user.

Abstract: Techniques are described herein that are capable of dynamically re-allocating computing resources while maintaining network connection(s). Applications of users are run in a computing unit. Computing resources are allocated among the applications based at least in part on dynamic demands of the applications for the computing resources and resource limits associated with the respective customers. In a first example, the computing resources are dynamically re-allocated among the applications, as a result of changing the resource limit of at least one customer, while maintaining at least one network connection between a client device of each customer and at least one respective application. In a second example, the computing resources are dynamically re-allocated among the applications, as a result of changing the resource limit of at least one customer, while maintaining at least one network connection between an interface and a client device of each customer.

Abstract: Example aspects include techniques for employing cross-service diagnostics for cloud service providers. These techniques may include dynamically generating a workflow of one or more diagnostic modules based on relationship information between an origin service experiencing an incident and one or more related services that the origin service depends on, and executing the workflow of one or more diagnostic modules to determine a root cause of the incident, each of the one or more diagnostic modules implemented by an individual service of the one or more related services in accordance with a schema. In addition, the techniques may include determining a diagnostic action based on the root cause, and transmitting, based on the diagnostic action, an engagement notification to a responsible entity.