Abstract: A first security rule set for traffic of a compute instance run at a virtualization server is stored in response to a programmatic request. A second security rule set for traffic of a network function accelerator of the virtualization server is stored in response to another programmatic request. Prior to delivery of network messages to the compute instance or from the accelerator, compliance with the applicable security rule set is verified.

Abstract: A system configured to improve a voice quality during a communication session by performing bandwidth extension on a narrowband speech signal to generate a wideband speech signal with higher audio quality. For example, a system can extend a speech bandwidth from a narrowband signal having a first bandwidth (e.g., 4 kHz) to a wideband signal having a second bandwidth (e.g., 8 kHz or higher). To perform bandwidth extension, the system may include cascaded neural networks, such as two or more sub-pixel convolutional neural networks (CNNs) connected in series. In some examples, a first sub-pixel CNN may extend the speech bandwidth from 4 kHz to 6 kHz and a second sub-pixel CNN may extend the speech bandwidth from 6 kHz to 8 kHz. Alternatively, the system may use three or more cascaded neural networks and/or may extend the speech bandwidth above 8 kHz without departing from the disclosure.

Abstract: Systems, methods, and computer-readable media are disclosed for modeled advertisement conversion attributions. An example method may include receiving first input data comprising first advertisement impression data and first advertisement conversion data, wherein the first input data includes one or more user identifiers associated with both the advertisement impression data and advertisement conversion data. The example method may also include training one or more machine learning models using the first input data. The example method may also include receiving second input data comprising second advertisement impression data, wherein user identifiers are unavailable for the second input data. The example method may also include determining, using the one or more machine learning models, second predicted conversion data associated with the second input data.

Abstract: Hop counts can be determined in a variety of network segments using packet headers collected at boundaries of the segments. Agents executing on network devices are used to transmit packet header data, including a TTL budget, to a collector server computer. The collector server computer can discern signal (production flows) from noise (traceroutes and probing traffic), detecting packets that are at risk of being dropped due to TTL expiration. Alerts can be generated for packet flows with a dangerously low remaining TTL budget, which are at high risk of expiring due to operational events resulting in traffic temporarily traversing slightly longer paths. Source addresses can be identified for such packet flows and network devices through which the packet flows are traversing can be identified. Hop counts can then be computed for different network segments, such as an Internet segment, a backbone segment and a data center segment.

Abstract: A detection mechanism is disclosed for identifying peers transmitting network traffic towards a cloud provider using routing prefixes that were not advertised for that peer. Once detected, an alarm condition can occur and the peer notified to remove any static route configurations. In one embodiment, a collector server computer can be used to acquire routing prefixes advertised to a peer. A second server computer can be used to obtain network traffic received from the peer. A comparison can then be made between the advertised prefixes and the network traffic being received. If network traffic is being received for unadvertised prefixes, then an alarm condition can occur. The detection mechanism can regularly inspect active traffic flows against BGP route advertisements at fixed intervals. For peers that are identified that are not in compliance with route advertisements, the alarm condition can occur for follow-up remediation.

Abstract: At an artificial intelligence system, training iterations of a first machine learning model are implemented. In a particular iteration, a group of data items are selected from an item collection using active learning, and respective labels selected from a set of tags are obtained for at least some of the items of the group. Using feature processing elements of a different machine learning model, a respective feature set corresponding to individual labeled items is generated in the iteration, and the feature sets are included in a training set used to train the first machine learning model. A trained version of the first machine learning model is stored after a training completion criterion is met.

Abstract: To achieve impact-less re-routing of traffic towards non best-paths in a network, two locator routes are originated for each network fabric: a primary locator and a secondary locator. In a steady state, the forwarding state on each router is created by recursion of endpoints identifiers through their originating location's primary locator-route. Secondary locator routes are not used in the steady state. At a desired point, a steering route is injected into the network device to redirect routing from the primary locator route to the secondary locator route. The secondary locator route has a different resultant port from the primary locator route to redirect incoming packets on a different network path.

Abstract: A candidate attribute combination of a first data set is identified, such that the candidate attribute combination meets a data type similarity criterion with respect to a collection of data types of sensitive information for which the first data set is to be analyzed. A collection of input features is generated for a machine learning model from the candidate attribute combination, including at least one feature indicative of a statistical relationship between the values of the candidate attribute combination and a second data set. An indication of a predicted probability of a presence of sensitive information in the first data set is obtained using the machine learning model.

Abstract: Disclosed are systems and methods that determine a likelihood that an item page represents a risky item. For example, a machine learning model may be trained to determine a probability score that an item page corresponds to a topic of a plurality of topics based on item textual descriptions determined from the item page and search queries corresponding to the item page. Still further, one or more of a topic risk score, item risk score, cumulative item risk score, and/or a total item risk score may be determined. Each of the risk scores may be indicative of a likelihood that the item page represents a risky item.

Abstract: Techniques for anomaly detection are described. An exemplary method includes receiving one or more requests to train an anomaly detection machine learning model using feedback-based training, the request to indicate one or more of a type of analysis to perform, a model selection indication, and a configuration for a training dataset; training the anomaly detection machine learning model according to the one or more requests using the training data; performing feedback-based training on the trained anomaly detection machine learning model; and using the retrained anomaly detection machine learning model.

Abstract: Techniques to reduce overhead in a direct memory access (DMA) engine can include processing descriptors from a descriptor queue to obtain a striding configuration to generate tensorized memory descriptors. The striding configuration can include, for each striding dimension, a stride and a repetition number indicating a number of times to repeat striding in the corresponding striding dimension. One or more sets of tensorized memory descriptors can be generated based on the striding configuration. Data transfers are then performed based on the generated tensorized memory descriptors.

Abstract: Execution status of managed time series processing tasks may be tracked. Status of a time series processing task that operations on different portions of a time series may be respectively captured. A request for the status of one of the portions of the time series with respect to the time series processing task may be received. The status may be identified and returned. For failed tasks, a failure reason may be generated by the time series processing system and included in a response with a failure status.

Abstract: A hybrid Bacon-Shor surface code is implemented using a fault tolerant quantum computer comprising hybrid acoustic-electric qubits. A control circuit includes an asymmetrically threaded superconducting quantum interference devices (ATS) that excites phonons in a mechanical resonator by driving a storage mode of the mechanical resonator and dissipates phonons from the mechanical resonator via an open transmission line coupled to the control circuit. The hybrid Bacon-Shor surface code only couples four phononic modes per given ATS, reducing cross-talk as compared to other systems that couple more phononic modes per ATS. Also, measurements are performed such that three parity measurements are taken between a phononic readout mode and a transmon qubit in a given syndrome measurement cycle.

Abstract: An apparatus powered by batteries may include a first processor and a second processor. The first processor may consume more electrical power than the second processor. The second processor acquires data from one or more sensors. At a predetermined time or other trigger, the second processor activates the first processor that is then used to send the acquired data to an external device. For example, items are stowed on the apparatus that includes weight sensors. The second processor awakens hourly, acquires weight data indicative of the weight of items stowed on the apparatus, and returns to a sleep mode. After 24 hours of acquiring weight data, the second processor activates the first processor to send the accumulated weight data to an external device, such as a server, using a network interface. Power consumption is substantially reduced, allowing the apparatus to operate without external power for extended periods of times.

Abstract: Movement of an autonomously motile device may be controlled by a user device. The user device may display image data captured by a camera of the autonomously motile device; a user may provide input, such as a touch gesture on a display screen, indicating a command for the autonomously motile device to move to a location indicated by the input. The autonomously motile device determines a coordinate of the input and a time of the touch input; the autonomously motile device then determines a direction and distance of a corresponding movement.

Abstract: Techniques for presence-detection devices to vary presence-detection sensitivity to detect different types of movements by objects, such as major movements and minor movements, using ultrasonic signals. The devices detect movement of a person in an environment by emitting ultrasonic signals into the environment, and characterizing the change in the frequency, or the Doppler shift, of the reflections of the ultrasonic signals off the person caused by the movement of the person relative to the presence-detection devices. The presence-detection devices may control the presence-detection sensitivity in order to detect major movements, such as a user walking in a room, as well as objects minor movements, such as a user typing at a computer. By adjusting the presence-detection sensitivity, the devices are able to improve the overall accuracy of detecting the presence of users in an environment.

Abstract: Disclosed are various embodiments for automated testing of systems and applications. A first computing device may be in data communication with a second computing device. The first computing device can receive a fault instruction specifying receive a fault instruction specifying a fault to inject into a virtual compute instance, parameters for the fault, a duration of the fault, and an identifier of the virtual compute instance. The first computing device can then send a command to a second computing device that hosts the virtual compute instance, wherein the command specifies at least the fault, the parameters for the fault, and the virtual compute instance. In turn, the second computing device can receive the command from the first computing device, save a pre-fault state of the virtual compute instance, and then introduce the fault into the virtual compute instance.

Abstract: Systems and methods are provided for dynamic segmentation of users during an experiment based on changes to application data collected during the experiment. Data regarding application interactions and associated application metadata may be collected from users during application experiments that involve testing different variants of a feature or otherwise different user experiences. The data regarding application interactions and associated application metadata may be evaluated to discover segments of users and/or usage patterns (e.g., “cohorts”). During the experiment, the users may be dynamically re-segmented into new/different cohorts based on new application data being collected.

Abstract: Representations of sets of descriptors of reference objects are stored in a repository, with individual descriptors including information about entities identified in the reference objects. In response to a request to extract content from a particular data object, a reference object which satisfies a similarity criterion with respect to the particular data object is identified from the repository using the descriptors. A structural comparison of the particular data object and the reference object is performed to determine an entity related to another entity identified in the particular data object.

Abstract: Descriptors of machine learning tasks to be used to respond to analysis requests, indicating acceptable categories of runtime environments for the tasks and metrics to be collected from the tasks, are received via programmatic interfaces. In response to an analysis request, an orchestrator receives results from individual tasks as they become available, provides the results to other tasks, and causes a response to the request to be prepared using results from at least a subset of the tasks. Metrics collected from the tasks, and a visual representation of the tasks indicating their runtime environments are presented.

Abstract: A service implemented at a first isolated virtual network of a provider network is added to a database of privately-accessible services. Configuration changes that enable network packets to flow between the first isolated virtual network and a second isolated virtual network without utilizing a network address accessible from the public Internet are implemented. Service requests originating at the second isolated virtual network are transmitted to the first isolated virtual network via private pathways of the provider network. Metrics corresponding to service requests directed from the second isolated network to the service are collected and provided to the respective owners of one or both isolated virtual networks.

Abstract: For a given cross-data-store transaction request at a storage service, a coordinator transmits respective voting transition requests to a plurality of log-based transaction managers (LTMs) configured for the respective data stores to which writes are directed in the transaction. The LTMs transmit responses to the coordinator based on data-store-specific conflict detection performed using contents of the voting transition requests and respective data-store-specific state transition logs. The coordinator determines a termination status of the cross-data-store transaction based on the LTMs' responses, and provides an indication of the termination status to the LTMs.

Abstract: Methods, systems, and computer-readable media for read-only data store replication to edge locations are disclosed. A read-only data store is replicated to an edge server. A request for content is received at the edge server from a client device. The content is sought from a content cache at the edge server or from an origin server coupled to the edge server. Processing of the request is initiated, comprising encountering an event. The event is associated with a function specified by a customer, and the function specifies the read-only data store as input. The function associated with the event is executed at the edge server, including retrieval of data from the data store. The content is generated based at least in part on execution of the function. The content is sent from the edge server to the client device.

Abstract: A network-accessible machine learning service is provided herein. For example, the network-accessible machine learning service provider can operate one or more physical computing devices accessible to user devices via a network. These physical computing device(s) can host virtual machine instances that are configured to train machine learning models using training data referenced by a user device. These physical computing device(s) can further host virtual machine instances that are configured to execute trained machine learning models in response to user-provided inputs, generating outputs that are stored and/or transmitted to user devices via the network.

Abstract: A fiber optic alignment assembly can include first and second receivers positioned on a base. The first receiver can include a channel for receiving a collet of a first fiber optic ribbon such that the ferrule of the first fiber optic ribbon is positioned for engagement. The second receiver can include a channel for receiving a collet of a second fiber optic ribbon such that a ferrule of the second fiber optic ribbon is positioned for engagement with the ferrule of the first fiber optic ribbon.

Abstract: A resource tagging service allows resource tags to be created and associated with computing resources in a service provider network. The resource tags can be utilized to search, collect, filter, organize and otherwise manage computing resources in the service provider network having matching tags, and/or to perform other types of functionality. A resource groups service allows customers and/or other users of the service provider network to view and access collections of computing resources that share common resource tags and/or other attributes. Resource groups can also be evaluated in order to identify computing resources in a service provider network for which certain types of actions or other functionality is to be performed. Membership of resource groups can also be evaluated over time and inferences can be drawn from the membership and from operational information associated with the member computing resources.

Abstract: A media marker mechanism may be used to send, to a cloud service, up-to-date context regarding playback of a media content stream on a user device. The cloud service may insert a media content item into a media content stream and/or combine media content items into a media content stream. The cloud service may implement the media marker mechanism to tag a content item with metadata that can be read by the user device. The user device can play the streaming media content and, when the tagged content item plays, read the metadata and send it to the cloud service. The cloud service can use the metadata to enrich the media content delivery by, for example, providing a companion image that the user device can display while paying the tagged content item, providing a link to make the companion image clickable, handling requests referring to the media content, etc.

Abstract: Described is a method for determining when to transition devices between different states based on determined potential activity in an area near the devices. When the potential for activity near an input device is below a threshold, the device is maintained in an inactive state. When the potential for activity near the device exceeds the threshold, the device is transitioned to an active state before the potential activity can occur near the device. Likewise, when the device is in an active state, the implementations described herein provide the ability to determine when to begin and end transmission of data from the device to a remote computing resource for processing. For example, obtained video may be processed locally to determine if an activity is occurring. If an activity is occurring, relevant data is sent to a remote computing resource to determine if an action has been performed during the activity.

Abstract: A system, method and processor that mitigates security vulnerabilities using context-dependent address space hiding. In some embodiments, a hardware mechanism allows a more-privileged software component managing multiple less-privileged software components to blind itself against “out-of-context” less-privileged software components. The hardware mechanism can allow the more-privileged software component to dynamically hide a portion of the more-privileged address space related to the “out-of-context” less-privileged software components, based on knowledge of the “in-context” less-privileged software component. A context register is set with a value from which an address range, within the address space of the more-privileged software component, can be determined, where the address range is associated with a first less-privileged software component can be determined.

Abstract: A determination is made that an explanatory data set for a common set of predictions generated by a machine learning model for records containing text tokens is to be provided. Respective groups of related tokens are identified from the text attributes of the records, and record-level prediction influence scores are generated for the token groups. An aggregate prediction influence score is generated for at least some of the token groups from the record-level scores, and an explanatory data set based on the aggregate scores is presented.

Abstract: A quantum computing service may store, in a cache, one or more compiled files of respective quantum functions included in one or more quantum computing programs received one or more customers. When the quantum computing service receives another quantum computing program, from the same or a different customer, the quantum computing service may determine whether the quantum computing program may include one or more of the quantum functions corresponding to the compiled files in the cache. If so, the quantum computing service may use the compiled files in the cache to compile the quantum computing program.

Abstract: A speech-processing system receives input data representing text. One or more encoders trained to predict audio properties corresponding to the text process the text to predict those properties. A speech decoder processes phoneme embeddings as well as the predicted properties to create data representing synthesized speech.

Abstract: Dialog acts (e.g., questions) are selected for voice browsing by a model trained to identify a dialog act that is most likely to lead to a desired outcome. Upon receiving an invocation to begin a conversation, a score indicative of a level of confidence that the conversation will have a successful outcome is determined, and a dialog act is selected based on the score. Subsequently, at each turn of the conversation, the score is updated or a new score is calculated, and a dialog act is selected based on the updated or new score. Confidence scores are calculated based on input features that are determined based on the user who uttered the invocation or responses to dialog acts, as well as a context of the conversation, and provided to a linear model or a machine learning model as inputs.

Abstract: Systems and methods are provided for presenting selectable page navigation options representing topics or search queries determined to be relevant to an item that a user is viewing on an initial page. When a user selects to begin entering a search query in a search filed on the initial page, a system may determine two or more page destinations to recommend to the user based on the first item presented on the initial page, where the page destinations may each represent a different topic or search query. These recommended destinations may then be presented over at least a portion of the initial page near the search field as selectable options for navigating to a corresponding recommended page destination. The selectable options may include, for each recommended page destination, an image representing the page destination and text identifying the page destination.

Abstract: Devices and techniques are generally described for a speech processing routing architecture. First input data representing an input request may be received. First data including a semantic interpretation of the input request may be determined. Metadata of the first input data may be determined. The metadata may identify an entity associated with the input request. In some examples, a query may be sent to a first component. The query may include the metadata. In some examples, second data that identifies a first skill associated with the entity may be received from the first component. In various examples, the first skill may be selected for processing the first input data based at least in part on the first data and the second data.

Abstract: A method includes extracting, using a backbone of a machine learning model, a plurality of features from an image of a foot and predicting, using a first portion of the machine learning model and based on one or more features of the plurality of features, a first aspect of the foot. The method also includes predicting, using a second portion of the machine learning model and based on one or more features of the plurality of features, a second aspect of the foot different from the first aspect, generating, using at least the first aspect and the second aspect, a two-dimensional model of a shoe, and superimposing the two-dimensional model of the shoe onto the image of the foot.

Abstract: Devices and techniques are generally described for learning personalized concepts for natural language processing. In various examples, a first natural language input may be received. In some examples, a determination may be made that the first natural language input comprises non-actionable slot data. A dialog session may be initiated with the user. In some examples, first slot data that is indicated by the user during the dialog session may be determined. In various examples, data representing the first slot data may be stored in a database in association with the first natural language input.

Abstract: A speech recognition system utilizing automatic speech recognition techniques such as end-pointing techniques in conjunction with beamforming and/or signal processing to isolate speech from one or more speaking users from multiple received audio signals and to detect the beginning and/or end of the speech based at least in part on the isolation. Audio capture devices such as microphones may be arranged in a beamforming array to receive the multiple audio signals. Multiple audio sources including speech may be identified in different beams and processed.

Abstract: Techniques for updating a machine learning (ML) model are described. A device or system may receive input data corresponding to a natural or non-natural language (e.g., gesture) input. Using a first ML model, the device or system may determine the input data corresponds to a data category of a plurality of data categories. Based on the data category, the device or system may select a ML training type from among a plurality of ML training types. Using the input data, the device or system may perform the selected ML training type with respect to a runtime ML model to generate an updated ML model.

Abstract: A database stores a document as a plurality of encrypted records, where each record is indicative of an incremental change to the state of the document, and encrypted using a document key. The document key is stored with encryption decryptable using a group key, and the group key is stored with encryption decryptable using a first access key. In response to a request to rotate from the first access key to a second access key, the database decrypts the group key using the first access key, a stores a group key re-encrypted with the second access key.

Abstract: Techniques for processing input data for a detected user are described. Received image data is processed to identify an indicated user. Based on the user a machine learning model is implemented. The machine learning model is then used to process input data for a user input. An action is performed using the resulting output data.

Abstract: A cable guide can include a first bracket and a second bracket. The first bracket can be configured to be coupled with a reference body The second bracket can be configured to be coupled with a moveable body moveable relative to the reference body. For example, the moveable body may correspond to a slidable server or other appliance receivable in a rack, and the reference body may correspond to a tray in the rack. A conduit structure may extend between the first bracket and the second bracket and include a first conduit laterally joined along a juncture to a second conduit. The first conduit and the second conduit may be separately serviceable. For example, the first conduit and the second conduit may be at least one of releasably separable along the juncture or separately openable along opposite sidewalls of the conduit structure.

Abstract: Techniques for low latency updating of an electrophoretic panel are described. According to such techniques, pixels may be continuously scanned out instead of buffering and updating at fixed scanning slots. In an example, a controller of a device loads, before receiving update data associated with a pixel of an electrophoretic panel, a LUT into a memory accessible to a LUT hardware module. After the LUT is loaded into the memory, the controller receives the update data indicating a color change to the pixel and associates the update data with the LUT. Based at least in part on the update data and the LUT, the controller determines waveform data corresponding to the color change and causes an output on the electrophoretic panel based at least in part on the waveform data.

Abstract: Methods, systems, and computer-readable media for automatic archiving of data store log data are disclosed. One or more operation records in a log are selected for archival. The one or more operation records comprise data indicative of operations performed on one or more data objects of a data store. The one or more operation records are selected for archival prior to deletion from the log. The one or more operation records are replicated from the log to an archive. Based at least in part on the replicating, the one or more operation records in the log are marked as archived. Based at least in part on the marking as archived, the deletion of the one or more operation records from the log is permitted.

Abstract: A new type system may be added to a type registry for a data processing service. A request to add the new type system may be received that describes the new type system for a data store. The new type system may be used to perform a data processing job that accesses the data store to obtain or store data as a source or target data store.

Abstract: Systems and methods are provided for management of data access latency in a distributed storage system by dynamically replicating data objects experiencing a high request volume. When a storage node is unable to meet a data access latency target (e.g., due to one or more data objects hosted by the storage node being the subject of a large quantity of requests), the storage node may initiate a process by which a second storage node is employed to reduce the request processing load. The second storage node may cache a copy of the requested data object, and may service at least a portion of requests for the data object using the cached copy.