Abstract: In a memory controller, a prefetch indication can be sent to memory to prepare the memory for a potential future read or write. Statistics can be used to select when such a prefetch should occur. The prefetch can occur without any read or write command being commenced. As a result, the memory controller predicts when to perform the prefetch. Some examples of when a prefetch can be sent include when there are other requests for the same memory page, or how often the page is requested. The page can remain open to prevent it from closing until the relevant read or write arrives. In the case that a read or write does not occur after a predetermined period of time, then a precharge can be performed to release the memory page.

Abstract: In one embodiment, a shelving unit has first and second sides, a frame, and a barrier. The frame defines a storage bay between the first and second sides, a first opening at the first side, and a second opening at the second side. The second opening receives a storage bin therethrough and into the storage bay. The first opening receives inventory items therethrough and into the storage bin when the storage bin is disposed in the storage bay. The barrier moves between (i) an open position, where the first opening is open to the storage bay so as to allow inventory items to be passed therethrough and into the storage bin, and (ii) a closed position, where the barrier at least partially obstructs at least one of the first and second openings so as to impede a human from reaching through the storage bay when the storage bin is removed.

Abstract: Techniques and systems for storing and retrieving data storage devices of a data storage system are disclosed. In some embodiments, inventory holders are used to store data storage devices used by a data storage system. When data is to be transacted with the data storage devices, mobile drive units locate appropriate inventory holders and transport them to a device reading station, where an appropriate device retrieval unit transacts the data. After the data has been transacted, the data storage devices are returned to the appropriate inventory holders, and the inventory holders are placed by the mobile drive units in locations where they may be accessed in response to further data transactions.

Abstract: Consistent distributed edge models via controlled dropout model training are described. According to some embodiments, a machine learning model is trained using multiple sensor data streams. During the training process, ones of the sensor data streams are dropped to cause the model to be generated to be robust in that it can tolerate missing input data from sensor data sources yet still maintain high accuracy. The model can be deployed to multiple sensor devices within an environment. The sensor devices generate sensor data and exchange a variety of types of data to ultimately result in a distributed, consistent model result being generated that remains accurate despite communication faults that may occur between ones of the sensor devices.

Abstract: Poses or gestures of actors within a scene may be detected and tracked using multiple imaging devices aligned with fields of view that overlap at least in part. Images captured by the imaging devices may be synchronized and provided to a classifier to recognize body parts within the images, and score maps indicative of locations of peak probabilities that the images include the respective body parts may be generated. Locations of peak values within the score maps may be correlated with one another to confirm that a given body part is depicted in two or more fields of view, and vectors indicative of distances to or ranges of motion of body parts, with respect to the given body part, may be generated. Motion of the body parts may be tracked in subsequent images, and a virtual model of the body parts may be generated and updated based on the motion.

Abstract: Embodiments are disclosed for launching a virtual machine and then configuring the virtual machine after launch. Metadata configuration information can be stored in conjunction with virtual machine image configuration data. The virtual machine image configuration data can be used to launch a virtual machine. The metadata configuration information can then be used to further configure the virtual machine after launch. Both the metadata configuration information and the virtual machine image configuration data can be stored in the same virtual machine registration record.

Abstract: Various embodiments of a system and method for tracking service requests are described. Embodiments may include call tree generation logic configured to receive multiple request identifiers associated with a respective one of multiple service requests. Each given request identifier may include an origin identifier, a depth value, and a request stack comprising one or more interaction identifiers. The call tree generation logic may also be configured to, based on multiple request identifiers that each include an origin identifier associated with a particular root request, generating a data structure that specifies a hierarchy of services called to fulfill that particular root request. Based on one or more of the interaction identifiers and one or more of the depth values, the generated data structure may specify for each given service of the hierarchy: a parent service that called the given service, and one or more child services called by the given service.

Abstract: Respective statistical distributions of a target variable within a proposed training data set and a proposed test data set for a machine learning model are obtained. A metric indicative of the difference between the two statistical distributions is computed. The difference metric is used to determine whether the proposed test data set is acceptable to evaluate the machine learning model.

Abstract: Techniques are described for providing users with access to computer networks, such as to enable users to interact with a remote configurable network service in order to create and configure computer networks that are provided by the configurable network service for use by the users. Computer networks provided by the configurable network service may be configured to be private computer networks that are accessible only by the users who create them, and may each be created and configured by a client of the configurable network service to be an extension to an existing computer network of the client, such as a private computer network extension to an existing private computer network of the client. If so, secure private access between an existing computer network and new computer network extension that is being provided may be enabled using one or more VPN connections or other private access mechanisms.

Abstract: Method and apparatus for stress management in a searchable data service. The searchable data service may provide a searchable index to a backend data store, and an interface to build and query the searchable index, that enables client applications to search for and retrieve locators for stored entities in the backend data store. Embodiments of the searchable data service may implement a distributed stress management mechanism that may provide functionality including, but not limited to, the automated monitoring of critical resources, analysis of resource usage, and decisions on and performance of actions to keep resource usage within comfort zones. In one embodiment, in response to usage of a particular resource being detected as out of the comfort zone on a node, an action may be performed to transfer at least part of the resource usage for the local resource to another node that provides a similar resource.

Abstract: Techniques for packaging and deploying algorithms utilizing containers for flexible machine learning are described. In some embodiments, users can create or utilize simple containers adhering to a specification of a machine learning service in a provider network, where the containers include code for how a machine learning model is to be trained and/or executed. The machine learning service can automatically train a model and/or host a model using the containers. The containers can use a wide variety of algorithms and use a variety of types of languages, libraries, data types, etc. Users can thus implement machine learning training and/or hosting with extremely minimal knowledge of how the overall training and/or hosting is actually performed.

Abstract: Features are disclosed for detecting words in audio using contextual information in addition to automatic speech recognition results. A detection model can be generated and used to determine whether a particular word, such as a keyword or “wake word,” has been uttered. The detection model can operate on features derived from an audio signal, contextual information associated with generation of the audio signal, and the like. In some embodiments, the detection model can be customized for particular users or groups of users based usage patterns associated with the users.

Abstract: In some embodiments, a system includes a proxy and an application server. The proxy is operable to intercept a request to download application content from an application server, insert an address of a navigation service into the request to yield an augmented request, and transmit the augmented request to the application server. The application server is operable to receive the augmented request, generate an application in response to the augmented request, the application comprising the application content, a navigation placeholder, and code to access the navigation service at the address, the navigation placeholder representing a portion of the application where the navigation service may be inserted, and transmit the application for receipt by the client.

Abstract: Techniques are described for verifying that an audio file corresponds to an instance of media content. An audio file is divided into a plurality of audio segments, and a digital fingerprint is generated for each of the plurality of audio segments. A digital signature is generated for the audio file by aggregating the digital fingerprints. The generated digital signature and at least one other digital signature corresponding to an instance of media content are processed as inputs to a linear regression machine learning model, to determine a measure of similarity between the generated digital signature and the at least one other digital signature. The linear regression machine learning model can be trained using a supervised learning approach and a set of training data. Embodiments determine whether the audio file corresponds to the instance of media content, based on the measure of similarity.

Abstract: A technology for job processing using a public service network. A method may include identifying processing availability at a public service network for processing a job submitted to a private service network. Available network bandwidth may be determined between the private service network and the public service network used to communicate between the private service network and the public service network and to transfer the job to the public service network for processing. Rules for transferring the job to the public service network may be identified. A determination may then be made that the processing availability at the public service network, the available network bandwidth between the private service network and the public service network, and the rules for transferring the job to the public service network allow the job to be transferred to the public service network for processing.

Abstract: Techniques for dynamically maintaining speech processing data on a local device for frequently input commands are described. One or more devices receive speech processing data specific to one or more commands associated with system input frequencies satisfying an input frequency threshold. The device(s) then receives input audio corresponding to an utterance and generate input audio data corresponding thereto. The device(s) performs speech recognition processing on input audio data to generate input text data using a portion of the received speech processing data. The device(s) determines a probability score associated with the input text data and determines the probability score satisfies a threshold probability score. The device(s) then performs natural language processing on the input text data to determine the command using a portion of the speech processing data. The device(s) then outputs audio data responsive to the command.

Abstract: At a network manager of an extension resource group of a provider network, a message comprising a command to launch a compute instance is received at an address which is part of a first network configured at a premise external to the provider network. The extension resource group includes a first host at the external premise. Within a second network configured at the external premise, the first host is assigned an address within a second address range. Addresses within the second range are also assigned to hosts within the provider network. The command is transmitted to the first host, and a compute instance is instantiated.

Abstract: Techniques for processing data according to customer-defined rules are disclosed. In particular, methods and systems for implementing a data alteration service using one or resources of a distributed computing system are described. The data alteration service is flexibly configurable by entities using the distributed computing system, and may be used to augment, compress, filter or otherwise modify data crossing a customer boundary.