Abstract: An encoding system that distributes a live stream to end user devices is provided herein. The encoding system automatically detects failed components and implements a failover action to replace the failed component with a backup component in a manner that reduces live stream interruptions. For example, the encoding system can include a network interface that is coupled to an encoder allocated to a live stream. Instead of providing a contribution encoder that transmits the live stream with a location of the allocated encoder, the encoding system can provide the contribution encoder with a location of the network interface. Thus, the contribution encoder can transmit the live stream to the network interface. The network interface can then forward the live stream to the allocated encoder. If the allocated encoder fails, then the network interface can forward the live stream to a backup encoder.

Abstract: Techniques are generally described for compression encoding of images. In various examples, the techniques may comprise receiving image data. In some examples, the techniques may further comprise generating a first compressed representation of luminance values of the image data by encoding luminance values of the image data with reference to a luminance palette. In some examples, the techniques may further include generating a second compressed representation of chrominance values of the image data by applying a frequency domain based image transform to the chrominance values of the image data. In various further examples, the techniques may further comprise combining the first compressed representation and the second compressed representation to generate compressed image data.

Abstract: A device for communicating including a housing including a camera, a microphone, a speaker, a button, a battery, a sensor, non-volatile memory, a processor, and a wireless communications module, wherein the non-volatile memory stores code operable by the processor for switching the processor from low-power mode to active mode in response to an activation trigger, receiving, from the one of the microphone and the camera, outbound audio and video signals, then sending a signal to a server via the wireless communications module during active mode, the signal including one or more of an alert signal, a signal based on the outbound audio signal, and a signal based on the outbound video signal, receiving from the server an inbound audio signal and outputting a signal based on the inbound audio signal via the speaker, and switching the processor from active mode to low-power mode in response to a deactivation trigger.

Abstract: A request to initiate a launch procedure of a compute instance at a virtualization host configured to access a remote storage device over a network is received. A memory buffer of the host is allocated as a write-back cache for use during a portion of the launch procedure. In response to a write request directed to remote storage during the portion of the launch procedure, the write payload is stored in the buffer and an indication of fulfillment of the write is provided independently of obtaining an acknowledgement that the payload has been propagated to the remote storage. Subsequent to the portion of the launch procedure, payloads of other write requests are transmitted to the remote storage device.

Abstract: An aggregator system obtains a first set of measurements from a logical partition of a computing resource monitoring service. The aggregator system aggregates this first set of measurements with a second set of measurements obtained from a storage resource. Using these measurements, the aggregator system generates a data segment that corresponds to a time interval of both the first set of measurements and the second set of measurements. The aggregator system stores this data segment in the storage resource and updates an index to indicate the location in the storage resource where the data segment is stored.

Abstract: A delivery autonomous ground vehicle includes several drainage features to keep water away from the package to be delivered and sensitive components. The cargo bay includes dimples to raise the package from the floor of the bay and a drain. Enclosed drain pans beneath elements of the lid also include drains. Water entering from the cooling air inlet holes of discharge holes can drain back through the holes or through the main drain. Overlapping panels inhibit water ingress, but if provided without seals, the water ingress between overlapping panels flows to the main drain.

Abstract: A mobile drive unit includes a pivot between the front chassis unit and the rear chassis unit, thereby diminishing the total height of the unit.

Abstract: A fixture includes a plurality of brackets. Accessories such as wire hooks, shelves, bins, and so forth may be supported by one or more brackets. A load cell supports each bracket and generates a signal that is indicative of changes to the bracket as items are added or removed from the accessory. The brackets include features such as tabs that prevent accessories from moving laterally. The fixtures are easily reconfigured to accommodate accessories of different sizes that in turn may support items of different sizes.

Abstract: Transitive restrictions can be applied to requests received on a session. A session token can be issued for an active session, and a transitivity setting specified to indicate the types of requests for which the transitive restriction is to be enforced. This can include enforcing the restriction on requests received from outside a trusted environment, requests within a scope of enforcement, or enforcing the restriction at request authentication. Any request received from an untrusted source that fails to satisfy the transitive restriction will be denied. Requests from inside the trusted environment may not have the transitive restriction enforced, such as where a new token is issued. This enables services within the environment to make calls on behalf of the customer, while ensuring that third parties obtaining the session token cannot successfully initiate such calls.

Abstract: The present disclosure provides for predictively allocating computing resources based on movement of a simulation agent through a simulation environment, such as in a video game. In order to support the movement of the simulation agent, computing resources may be allocated in various ways. For example, one option may be to transfer data assets associated with the simulation agent from the first computing resource to the second computing resource. Another option may be leave the data assets at the first computing resource and transfer authority over at least a portion of the second region from the second computing resource to the first computing resource. In various embodiments, the way in which the computing resources are allocated may be determined based on various factors, such as resource costs, so that the movement of the simulation agent can be supported with optimal system efficiency.

Abstract: In a computer comprising a plurality of integrated circuits (ICs), each IC may be connected to all other ICs via a respective point-to-point interconnect. A source IC divides the data to be transmitted to a destination IC for a transaction to generate multiple data cells so that each data cell includes a different portion of the data. The source IC transmits one of the data cells to the destination IC and remaining data cells to intermediate ICs, wherein an intermediate IC is an IC other than the source IC or the destination IC. The intermediate ICs forward the remaining data cells to the destination IC.

Abstract: Techniques for optimizing background tasks based on forecast data are described. Customer workloads may be monitored by a local monitor in a first time period. Future customer workloads in a second time period following the first time period may be forecast based at least on the customer workloads using a local model. A background availability may be determined based at least on the future customer workloads. Execution of at least one background workload may be scheduled to use the background availability during the second time period. The local monitor may then cause the execution of the at least one background workload.

Abstract: Scheduling of the operations of an integrated circuit device such as a hardware accelerator, including scheduling of movement of data into and out of the accelerator, can be performed by a compiler that produces program code for the accelerator. The compiler can produce a graph that represents operations to be performed by the accelerator. Using the graph, the compiler can determine estimated execution times for the operations represented by each node in the graph. The compiler can schedule operations by determining an estimated execution time for set of dependent operations that depend from an operation. The compiler can then select an operation that has a shortest estimated execution time from among a set of operations and which has a set of dependent operations that has a longest estimated execution time as compared to other sets of dependent operations.

Abstract: Categorizing images may include training a first neural network to cluster a plurality of images to obtain a first image embedding space, wherein a vector representation is determined for each of the plurality of images based on the training, determining a vector norm value corresponding to each of the plurality of images based on the vector representation for each of the plurality of images, and identifying a first subset of the images for which a corresponding vector norm value satisfies a predetermined vector norm quality threshold. Then, a second neural network may be trained using the first subset of images to obtain a second image embedding space, and the second image embedding space may be used to categorize additional images.

Abstract: A deployment subsystem provides updates to an application and/or software executed by server computer systems. The update is included in an update script. Execution of the update script by a server computer system causes the server computer system to execute a set of checks. After completing the set of checks the server computer system may execute the update to the application and/or software executed by server computer systems.

Abstract: Methods and apparatus are described by which media player functionality such as, for example, adaptive bitrate selection may be supported in media players on iOS, tvOS, or MacOS devices for video content encoded using the HEVC codec. Each video segment reference in the manifest includes two associated HLS-compliant tags, a discontinuity tag followed by a map segment tag.

Abstract: Disclosed are various embodiments for searchable encrypted data stores. A plurality of rows in a data table that have a cell in a column of the row that matches a value are identified. A bitmap that represents the plurality of rows is generated. The bitmap is encrypted. A key is then generated for a key-value pair. The key can be represented by a message authentication code (MAC) based at least in part on a tuple of a name of the data table, an identifier of the column, and the value. The key-value pair with the key and the encrypted bitmap is then sent to a remote computing device for storage.

Abstract: Techniques for detecting anomalies in streaming graph data are described. For example, an embedding technique of generating a multi-dimensional vector of summations of each weighted edge found in both a random source bounding proper subset and a random destination bounding proper subset associated with a dimension of the epoch graph is detailed. Anomaly detection is performed on the generated multi-dimensional vectors.

Abstract: Products such as food, clothing or machines may be customized for a customer based on any preferences that may be determined from information that is known, determined or predicted regarding the customer. Where a customer's browsing, purchasing or search histories, or other data, indicate that the customer prefers (or dislikes) products having a specific ingredient or component, a product may be customized to include (or exclude) the ingredient or component, and recommended to the customer. Customized products may include food, clothing, machines or any other articles that may be fabricated with or altered to include specific components. A customized product may be marketed to a customer in any manner, and may be prepared and delivered to the customer upon request.

Abstract: Disclosed are techniques regarding aspects of selectively enabling or disabling interconnects between peripheral processors. The peripheral processors may be included within a compute service provider and may be accessible via virtual machines within guest domains implemented within the compute service provider. The interconnects can be enabled or disabled depending upon whether they traverse domains of the compute service provider.

Abstract: A replica database may receive a replication instruction from a master. The replica may maintain a version of a collection of data in database pages stored on a storage device. A replication thread may process the instruction. The replication thread may determine that an item affected by the instruction is located in a page that is both loaded into a memory buffer and subject to contention. The page may be modified with information indicating that the instruction has not been processed. A subsequent reader thread may, while processing a request to access the page, apply the instruction and complete processing of the request.

Abstract: Accessing measurements associated with garments and leveraging the measurements to determine fit characteristics associated with the garments is described. The fit characteristics may be leveraged to provide information about how a garment fits relative to another garment to streamline online shopping transactions. Fit characteristics may correspond to scores that are determined based at least in part on one or more measurements of a garment. Individual measurements may be standardized and/or weighted. The individual measurements may be combined to determine a score representative of a fit characteristic. Fit characteristics associated with two or more garments may be compared and a user interface may be caused to be displayed to compare garments. The user interface may summarize fit characteristic comparisons via textual descriptions, graphical representations, etc. Users may leverage fit characteristics and resulting user interfaces to comparison shop between various garments.

Abstract: Disclosed are various embodiments for updating account data with multiple account providers. Account management logic determines that data associated with a user has been updated. A validation procedure is performed on the updated data. Multiple accounts of the user that may use the data are determined. The accounts are with multiple account providers. Corresponding account update requests for the accounts are sent to the account providers. The account update requests specify the data that has been updated.

Abstract: The present disclosure is directed to training, and providing recommendations via, a machine learning model architected to balance relevance and diversity of sets of recommendations. For example, a neural network can be provided with user profile features and can output probabilities for each of a number of recommendations. This can be converted into a ranked list of recommendations. The ranked list of recommendations is provided to a diversity model that maximizes an optimization objective having a first objective that quantifies relevance of a recommendation and a second objective that measures diversity of a set of recommendations. The output of the diversity model is a set of recommendations that have both high relevance and high diversity.

Abstract: Techniques for updating voice profiles used to perform user recognition are described. A system may use clustering techniques to update voice profiles. When the system receives audio data representing a spoken user input, the system may store the audio data. Periodically, the system may recall, from storage, audio data (representing previous user inputs). The system may identify clusters of the audio data, with each cluster including similar or identical speech characteristics. The system may determine a cluster is substantially similar to an existing voice profile. If this occurs, the system may create an updated voice profile using the original voice profile and the cluster of audio data.

Abstract: This patent application is directed to interaction between a voice-controlled device and a language processing system that processes user requests. The requests can be requests for information or they can be requests to control a device, such as “Turn on the bedroom light.” The disclosure includes various embodiments for trying to resolve system errors in a manner that provides an improved customer experience. The improvements described include the capability of the overall system to recognize that an error has occurred while attempting to fulfill a request that was submitted as part of a recognized spoken utterance. Instead of simply doing nothing, or of playing a message to the customer to try again later, the system would now let the user know it was still processing the request while making another attempt at fulfilling the request.

Abstract: Systems and methods are described for bootstrapping an application with metadata specified in a template. The template specifies a stack of resources that will be used to execute an application and also includes a set of metadata for customizing the resources and the application. When the system receives the template, it instantiates a compute node which will execute the application. The compute node may contain at least one initialization script for bootstrapping the application with the metadata contained in the template. This functionality allows users to bootstrap the application running on the compute node with data from within the template that was used to create the stack of resources for executing the application. In this manner, metadata, configuration files, package names and versions can be passed by the application owner to the remote compute node.

Abstract: A determination is made as to whether a value of a first parameter of a first application is to be obtained using a natural language interaction. Based on received input, a first service of a plurality of services is identified. The first service is to be used to perform a first task associated with the first parameter. Portions of the first application to determine the value of the first parameter and to invoke the first service are generated.

Abstract: Route tables may be associated with ingress traffic for logically isolated networks. A routing device at the edge of a logically isolated network may receive a route to include in a route table that is associated with ingress traffic to the logically isolated network to forward the ingress traffic to a network appliance hosted in the logically isolated network. Network packets received at the edge routing device may have a destination of a computing resource hosted in the logically isolated network. The edge routing device may identify the route in the route table to override the destination in the network packet with the network appliance and forward the network packet to the network appliance according to the route.

Abstract: A virtualization host is identified for an isolated run-time environment. One or more records generated at a security module of the host, which indicate that a first phase of a multi-phase establishment of an isolated run-time environment has been completed by a virtualization management component of the host, is transmitted to a resource verifier. In response to a host approval indicator from the resource verifier, the multi-phase establishment is completed at the virtualization host.

Abstract: Disclosed are methods and systems for delivery of items using an unmanned aerial vehicle (“UAV”). A user may be provided with a delivery location identifier (“DLI”) that is to be placed at a delivery location within a delivery destination to identify where a UAV is to position an item as part of a delivery to the delivery destination. For example, the delivery destination may be a user's home. Within the deliver destination of the user's home, the user may select a delivery location, such as a spot in the back yard wherein the UAV is to position the ordered item as part of the delivery. To aid the UAV in navigating to the delivery location, the user places the DLI at the delivery location. The UAV detects the DLI and positions the item at or near the DLI as part of the item delivery.

Abstract: New application sessions can be placed on reclaimable resource capacity that can be provided at lower cost than dedicated capacity, but with risk that the capacity might be reclaimed before completion of the session. A number of instance types can be determined that are capable of hosting a new session. Risk scores can be calculated for the instance types, as well as the cost of instances of each type. An instance type can be selected for the session using the risk scores and cost information, where the instance type will have relatively low risk with a relatively low cost for the instance. Once an instance type is selected, a resource pool can be selected from which to provide the resource instance of the selected type. The instance can then be allocated and caused to host the session or perform related functionality.

Abstract: The present disclosure is directed to generating neural network (NN) output using input data representing various types of events, such as input representing a certain type of event and also an engagement metric that may be representative of a property of the event or representative of a related but different type of event. For example, the output values generated using the NN may be associated with the likelihood that certain future events will occur, given the occurrence of certain past or current events. The output can then be modified (e.g., re-ranked, adjusted, etc.) based on the occurrence of certain other past or current events.

Abstract: Aerial vehicles are equipped with propellers having clutch mechanisms that contract around a shaft when the propellers are not rotating, or are rotating at low angular velocities, and expand around the shaft when the propellers are rotating at sufficiently high angular velocities. The clutch mechanisms surround one or more fixed posts within an opening or window defined therein. When the clutch mechanisms contract into a closed position, components of the clutch mechanisms come into contact with the posts, and the propellers are forced to remain in an alignment defined by the posts. When the clutch mechanisms expand into an open position, such components may rotate freely without contacting the posts. The clutch mechanisms cause propellers to remain aligned in desired orientations when the propellers are not required for operation, thereby reducing drag or adverse acoustic effects.

Abstract: Systems, methods, and computer-readable media are disclosed for removable conveyor cartridge assemblies for mobile carrier units. In one embodiment, an example removable assembly for a mobile carrier unit of a sortation system may include a frame, a conveyor belt disposed about the frame, a first sheet metal component coupled to a first side of the frame, the first sheet metal component including a first set of tabs extending from the first sheet metal component, and a tensioning device configured to tension the conveyor belt, the tensioning device having a first screw that engages a first tab of the first set of tabs.

Abstract: Item singulation systems and methods may include an input conveyor, a separation conveyor, and an output conveyor that is positioned orthogonally to the input conveyor. A vision system may determine individual columns of items from among a plurality of items on the input conveyor. Operations of the input conveyor, separation conveyor, and output conveyor may be controlled substantially as a pull system to transfer individual columns of items to the output conveyor, such that singulated items can be transferred by the output conveyor to downstream processes.

Abstract: A video may include a runtime. Based on analysis of the video, a scene or shot change may be generated based on the content of the video. For instance, the video may depict a character inside of a house in a first shot and, in a subsequent shot, the video may depict the character as being outside in a forest. A shot change may indicate the transition from the character being depicted inside in the first shot and outside in the second shot at a period of time during the runtime of the video. The video may be fragmented into fragments based on the shot change and based on a minimum or maximum fragment duration. When a video is fragmented into fragments, a key frame that is independently decodable may be inserted into a fragment. Encoding settings may then be determined based on the key frame.

Abstract: Techniques are described for using machine learning (ML) models to create information technology (IT) infrastructures at a service provider network based on image of IT system architecture diagrams. To create IT system architecture diagrams, system architects often use tools ranging from pen and paper and whiteboards to various types of software-based drawing programs. Based on a user-provided image of an IT system architecture diagram (for example, a digital scan of a hand drawn system diagram, an image file created by a software-based drawing program, or the like), a service provider network uses one or more ML models to analyze the image to identify the constituent elements of the depicted IT system architecture and to create an infrastructure template that can be used to automatically provision corresponding computing resources at the service provider network.

Abstract: A method comprising receiving subject data indicative of a subject entity and selecting, from a plurality of data sets, and based on the subject data, a subject entity data set which corresponds to the subject entity. The subject entity data set comprises first related entity data representative of a first related entity related to the subject entity and first text data representative of first text associated with the first related entity. Unstructured text data representative of unstructured text is received and processed, using the first text data, to identify a portion of the unstructured text data corresponding to the first text data. The first text data is used to identify, from the subject entity data set, the first related entity data and the portion of the unstructured text data is identified as corresponding to the first related entity data.

Abstract: A system operated by a computing resource service provider maintains a pool of computing resources for use in processing job requests submitted by customers. The system allocates computing resources to the pool in anticipation of future resource needs. In various embodiments, future resource needs can be estimated based on scheduled jobs, or historical job information. In an embodiment, the computing resources are virtual computer systems which may be arranged in a cluster. In response to receiving a parallel processing job from a customer, the system reserves computing resources from the pool for performing the job. In an embodiment, the reserved resources are configured with a network namespace that is able to access to a customer's resources.

Abstract: Updates performed as part of transaction requests to a data store may be included in a stream of updates. Updates to items in the data store that are included in transactions determined not to be committed to the data store may be excluded from the stream of updates. Records in the stream of updates may include an identifier for the transaction that included the update described by the record. The identifier for the transaction may be used to identify updates to other items in the data store that are included in the same transaction.

Abstract: A system and method of generating a cleaned machine image. One example includes capturing used machine image data stored in a first boot volume to generate a set of captured machine image data and extracting at least a portion of the set of captured machine image data to the second volume to generate a cleaned machine image.

Abstract: Multimedia content may be obtained and an object may be identified in a first frame of video content. The object may be tracked through a plurality of frames, and the object may be identified in a second frame of the video content only if the object is no longer substantially identifiable.

Abstract: A server computer toggles between a protected mode and an unprotected mode. In the protected mode, users are unable to access configuration information due to a Base Address Register (BAR) being cleared. However, a service provider can access a Trusted Platform Module (TPM) through an Application Program Interface (API) request. In an unprotected mode, the BAR is programmed so that users can access the configuration information, but the TPM is blocked. Blocking of the TPM is achieved by changing a configuration file, which changes an overall image of the card. With the modified image not matching an original image, the TPM blocks access to data, such as encryption keys. Separate interfaces can be used for user access (PCIe) and service provider access (Ethernet) to the server computer. The server computer can then be toggled back to the protected mode by switching the configuration file to the original configuration file.

Abstract: Systems, methods, and computer-readable media are disclosed for dynamically determining highlights of media content based on user interaction metrics and/or social media metrics. In one embodiment, an example method may include determining media content streamed to user devices over a time period and corresponding to a live event, determining user interaction data associated with the media content and indicative of user interactions with the user devices, determining, using the user interaction data, a quality score for a time interval during the time period, the quality score indicative of user engagement in a segment of the media content, determining, using the time interval and segment-by-segment metadata, a stream start time and a stream end time for the segment, generating, using the stream start time and the stream end time, a clip of the segment, and determining, using the quality score, that the clip is to be included in a highlight.

Abstract: Techniques for capturing spoken user inputs while a device is prevented from capturing such spoken user inputs are described. When a first device has a status representing it is unbeneficial for the first device to perform wakeword detection, a second device (e.g. a vehicle) may perform wakeword detection on behalf of the first device. The second device may be unable to send audio data, representing a spoken user input, to a speech processing system. In such an example, the second device may send the audio data to a third device, which may send the audio data to the speech processing system.

Abstract: Techniques for maintaining a graph associated with snapshots are described. In some embodiments, a snapshot service maintains one or more graphs in a graph database. The snapshot service maintains a graph data structure having a plurality of vertices corresponding to a plurality of snapshots, wherein each vertex of the plurality of vertices is at least a parent to or a child of another vertex, and each child vertex of the plurality of vertices comprises a value representing a number of chunks of data referenced in the corresponding snapshot corresponding to of the child vertex that are not referenced in the snapshot corresponding to the parent vertex of the child vertex. The snapshot service further determines a total number of unique chunks of data attributed to the plurality of snapshots based at least in part on traversing the graph data structure.

Abstract: In some examples, a movement range indicating a geospatial boundary for executing delivery of a package may be received by a delivery tag. The geospatial boundary may be stored by the delivery tag. A position of the delivery tag with respect to the geospatial boundary can be determined using a location sensor of the delivery tag. It may be detected that the delivery tag has moved outside the geospatial boundary, in response to which an alarm may be generated.

Abstract: Systems and methods are disclosed to implement a multi-region event publishing system to broadcast heterogeneous events across regions. In embodiments, an event publisher at a first location receives an event to be published and writes the event to a first data store of a multi-location replicated data storage system. The event publisher may generate metadata for the event including a source location, destination location(s), and a timestamp. The event is replicated to a second data store of the multi-location replicated data storage system at a second location. The second data store generates a change stream to an event receiver. The event receiver filters the second change stream based on the event metadata, and provides the filtered events to registered event consumers.

Abstract: An integrated circuit device such as a neural network accelerator can be programmed to select a numerical value based on a multinomial distribution. In various examples, the integrated circuit device can include an execution engine that includes multiple separate execution units. The multiple execution units can operate in parallel on different streams of data. For example, to make a selection based on a multinomial distribution, the execution units can be configured to perform cumulative sums on sets of numerical values, where the numerical values represent probabilities. In this example, to then obtain cumulative sums across the sets of numerical values, the largest values from the sets can be accumulated, and then added, in parallel to the sets. The resulting cumulative sum across all the numerical values can then be used to randomly select a specific index, which can provide a particular numerical value as the selected value.