Abstract: Edge devices of a network collect data. An edge device may determine whether to process the data using a local data processing model or to send the data to a tier device. The tier device may receive the data from the edge device and determine whether to process the data using a higher tier data processing model of the tier device. If the tier device determines to process the data, then the tier device processes the data using the higher tier data processing model, generates a result based on the processing, and sends the result to an endpoint (e.g., back to the edge device, to another tier device, or to a control device). If the tier device determines not to process the data, then the tier device may send the data on to another tier device for processing by another higher tier model.

Abstract: A data source configured to provide a representation of an environment of one or more agents is identified. Using a data set obtained from the data source, a neural network-based reinforcement learning model with one or more attention layers is trained. Importance indicators generated by the attention layers are used to identify actions to be initiated by an agent. A trained version of the model is stored.

Abstract: Various embodiments of a hierarchical system or method of identifying sensitive content in data is described. In some embodiments, sensitive data classifiers local to a data storage system can analyze a plurality of data items and classify at least some data items as potentially containing sensitive data. The sensitive data classifiers can provide the classified data items to a separate sensitive data discovery component. The sensitive data discovery component can, in some embodiments, obtain the classified data items, perform a sensitive data location analysis on the classified data items to identify a location of sensitive data within some of the classified data items, and generate location information for the sensitive data within the data items containing sensitive data. The sensitive data discovery component can provide to a destination this information, in some embodiments, where the destination might redact, tokenize, highlight, or perform other actions on the located sensitive data.

Abstract: A data source configured to provide a representation of an environment of one or more agents is identified. Using a data set obtained from the data source, a neural network-based reinforcement learning model with one or more attention layers is trained. Importance indicators generated by the attention layers are used to identify actions to be initiated by an agent. A trained version of the model is stored.

Abstract: Techniques for performing natural language processing (NLP) on semi-structured data are described. An exemplary method includes receiving a semi-structured document to perform NLP on using a trained NLP model; converting the semi-structured document into a secondary format, wherein the secondary format includes spatial information for tokens of the semi-structured document; flattening the converted, secondary formatted semi-structured document into a Unicode Transformation Format text file; performing NLP on the Unicode Transformation Format text file using the trained NLP model; and providing a result of the NLP to a requester.

Abstract: A simulation management service receives a request to perform reinforcement learning for a robotic device. The request can include computer-executable code defining a reinforcement function for training a reinforcement learning model for the robotic device. In response to the request, the simulation management service generates a simulation environment and injects the computer-executable code into a simulation application for the robotic device. Using the simulation application and the computer-executable code, the simulation management service performs the reinforcement learning within the simulation environment.

Abstract: An artificial intelligence (AI) system generates a self-contained AI container packaged as an executable that can be installed on client devices or computing infrastructure. The AI container envelopes a set of components capable of performing one or more processes related to machine learning operations, including data collection, data processing, feature engineering, data labelling, model design, model training and performance optimization for underlying hardware, model deployment, and model feedback monitoring, and the like in a local environment of a client device or cloud environment of a computing infrastructure. The AI system packages the self-contained AI container as an executable that can be installed on client devices of any hardware configuration or operating system.

Abstract: A training system may create and train a machine learning model with knowledge transfer. The knowledge transfer may transfer knowledge that is acquired by another machine learning model that has been previously trained to the machine learning model that is under training. The knowledge transfer may include a combination of representation transfer and instance transfer, the two of which may be performed alternatingly. The instance transfer may further include a filter mechanism to selectively identify instances with a satisfactory performance to implement the knowledge transfer.

Abstract: Indications of sample machine learning models which create synthetic content items are provided via programmatic interfaces. A representation of a synthetic content item produced by one of the sample models in response to input obtained from a client of a provider network is presented. In response to a request from the client, a machine learning model is trained to produce additional synthetic content items.

Abstract: A training system may create and train a machine learning model with knowledge transfer. The knowledge transfer may transfer knowledge that is acquired by another machine learning model that has been previously trained to the machine learning model that is under training. The knowledge transfer may include a combination of representation transfer and instance transfer, the two of which may be performed alternatingly. The instance transfer may further include a filter mechanism to selectively identify instances with a satisfactory performance to implement the knowledge transfer.

Abstract: Techniques for heterogeneous compute instance auto-scaling with reinforcement learning (RL) are described. A user specifies a reward function that generates rewards for use with an application simulation for determining what different instance types should be added to or removed from the application as part of training a RL model. The RL model can be automatically deployed and used to monitor an application to automatically scale the application fleet using heterogenous compute instances.

Abstract: Indications of sample machine learning models which create synthetic content items are provided via programmatic interfaces. A representation of a synthetic content item produced by one of the sample models in response to input obtained from a client of a provider network is presented. In response to a request from the client, a machine learning model is trained to produce additional synthetic content items.

Abstract: A simulation workflow manager obtains a set of parameters for simulation of a system and training of a reinforcement learning model for optimizing an application of the system. In response to obtaining the set of parameters, the simulation workflow manager configures a first compute node that includes a training application for training the reinforcement learning model. The simulation workflow manager also configures a second compute note with a simulation application to perform the simulation of the system in a simulation environment. Data is generated through execution of the simulation in the second compute node that is provided to the first compute node to cause the training application to use the data to train the reinforcement learning model.

Abstract: A hub device of a network receives data from edge devices and generates a local result. The hub device also sends the data to a remote provider network and receives a result from the remote provider network, wherein the result is based on the data received from the edge devices. The hub device then generates a response based on the local result or the received result. The hub device may determine to correct the local result based on the result received from the remote provider network, and generate the response based on the corrected result. The hub device may generate an initial response before receiving the result from the provider network. For example, the hub device may determine that the confidence level for the local result is above the threshold level and in response, generate the initial response based on the local result.

Abstract: Methods, systems, and computer-readable media for context-sensitive linking of entities to private databases are disclosed. An entity linking service stores a plurality of representations of entities. Individual ones of the entities correspond to individual ones of a plurality of records in one or more private databases. The entity linking service determines a mention of an entity in a document. The entity linking service selects, from the plurality of records in the one or more private databases, a record corresponding to the entity. The record is selected based at least in part on the plurality of representations of the entities and based at least in part on a context of the mention of the entity in the document. The entity linking service generates output comprising a reference to the selected record in the one or more private databases.

Abstract: Methods, systems, and computer-readable media for event extraction from documents with co-reference are disclosed. An event extraction service identifies one or more trigger groups in a document comprising text. An individual one of the trigger groups comprises one or more textual references to an occurrence of an event. The one or more trigger groups are associated with one or more semantic roles for entities. The event extraction service identifies one or more entity groups in the document. An individual one of the entity groups comprises one or more textual references to a real-world object. The event extraction service assigns one or more of the entity groups to one or more of the semantic roles. The event extraction service generates an output indicating the one or more trigger groups and one or more entity groups assigned to the semantic roles.

Abstract: Methods, systems, and computer-readable media for lifecycle management for customized natural language processing are disclosed. A natural language processing (NLP) customization service determines a task definition associated with an NLP model based (at least in part) on user input. The task definition comprises an indication of one or more tasks to be implemented using the NLP model and one or more requirements associated with use of the NLP model. The service determines the NLP model based (at least in part) on the task definition. The service trains the NLP model. The NLP model is used to perform inference for a plurality of input documents. The inference outputs a plurality of predictions based (at least in part) on the input documents. Inference data is collected based (at least in part) on the inference. The service generates a retrained NLP model based (at least in part) on the inference data.

Abstract: A simulation workflow manager obtains a set of parameters for simulation of a system and training of a reinforcement learning model for optimizing an application of the system. In response to obtaining the set of parameters, the simulation workflow manager configures a first compute node that includes a training application for training the reinforcement learning model. The simulation workflow manager also configures a second compute note with a simulation application to perform the simulation of the system in a simulation environment. Data is generated through execution of the simulation in the second compute node that is provided to the first compute node to cause the training application to use the data to train the reinforcement learning model.

Abstract: Systems and methods are described for providing an implementation of the MapReduce programming model utilizing tasks executing on an on-demand code execution system or other distributed code execution environment. A coordinator task may be used to obtain a request to process a set of data according to the implementation of the MapReduce programming model, to initiate executions of a map task to analyze that set of data, and to initiate executions of a reduce task to reduce outputs of the map task executions to a single results file. The coordinator task may be event-driven, such that it executes in response to completion of executions of the map task or reduce tasks, and can be halted or paused during those executions. Thus, the MapReduce programming model may be implemented without the use of a dedicated framework or infrastructure to manage map and reduce functions.

Abstract: A model training service of a provider network receives data from edge devices of a remote network. The model training service analyzes the received data. The model training service may also analyze global data from other edge devices of other remote networks. The model training service may then generate updates to local data processing models based on the analysis. The updates are configured to update the local data processing models at the edge devices of the remote network. The provider network deploys the updates to the remote network. The updates are then applied to the data processing models of the edge devices.