Abstract: In some embodiments, a first output is received from a first prediction network at a second prediction network. The first prediction network generates the first output from a first input. Also, a second input is received at the second prediction network that describes the first input. The second prediction network analyzes the first output and the second input and generates a second output that classifies the first output in one of a set of classifications. The first output is output with the one of the set of classifications for the second output where the second output indicates whether the first output should be reviewed when the second output is classified in a first classification in the set of classifications or not reviewed when the second output is classified in a second classification in the set of classifications.

Abstract: Distributed vector representations of source code commits, are generated to become part of a data corpus for machine learning (ML) for analyzing source code. The code commit is received, and time information is referenced to split the source code into pre-change source code and post-change source code. The pre-change source code is converted into a first code representation (e.g., based on a graph model), and the post-change source code into a second code representation. A first particle is generated from the first code representation, and a second particle is generated from the second code representation. The first particle and the second particle are compared to create a delta. The delta is transformed into a first commit vector by referencing an embedding matrix to numerically encode the first particle and the second particle. Following classification, the commit vector is stored in a data corpus for performing ML analysis upon source code.

Abstract: A first machine learning model may be trained to generate a first representation of a first user data having private attributes and non-private attributes. The first representation may be generated to achieve a focal entropy by maximizing an entropy amongst similar private attributes. The first representation to preserve information associated with the non-private attributes but omit information associated with the private attributes. Moreover, the first user data may be classified based on a target portion of the first representation including the non-private attributes but not the residual portion of the first representation including the private attributes. The trained first machine learning model may be applied to generate a second representation of a second user data such that downstream tasks may be performed by applying a second machine learning model to the second representation of the second user data. Related systems and computer program products are also provided.

Abstract: Techniques and solutions are described for implementing virtual domains. Computing resources in a computing environment are determined and assigned to one or more virtual domains. One or more data structures can be located in a given virtual domain. The computing resources assigned to a virtual domain can be dynamically reconfigured without affecting processes that submit tasks to be performed on data structures in the virtual domains. Tasks can be submitted to a dispatcher, which can determine the appropriate virtual domain for the task and forward the task to the determined virtual domain. Tasks are received by virtual domains and assigned to worker threads, which can access a data structure specified for a given task.

Abstract: A method can include receiving, at a node comprising a distributed trust data storage system, a request to execute a transaction modifying a portion of a data partition. Replicas of the data partition may be stored at a plurality of nodes comprising the distributed trust data storage system. In response to the request, the transaction can be sent to the plurality of nodes storing replicas of the data partition to modify the portion of the data partition. The transaction can be committed based on a threshold quantity of the plurality of nodes reaching a consensus by determining a same cryptographic hash for the transaction. The transaction can be committed by sending, to the plurality of nodes, an indication to add an entry corresponding to the transaction to a transaction log recording transactions executed on the portion of the data partition. Related systems and articles of manufacture are also provided.

Abstract: Configuration data characterizing configuration information of at least one application process executing on a target system is received. The configuration data includes at least one first table and at least one second table upon which the first table depends. The receiving is performed prior to transmission of the configuration data to the target system for application of the configuration data to the target system. At least one validation definition including a header and table elements specifying a dependency of the first table on the second table is determined using the configuration data. Whether to validate the configuration data to ensure consistency within the target system is determined using the at least one validation definition. The determining whether to validate includes determining to validate the configuration data in response to determining that the second table has been modified. Related apparatus, systems, techniques and articles are also described.

Abstract: In one embodiment, the present disclosure pertains to automated data type determination of variables that are written in a programming language. In one embodiment, a programming language statement is received. The programming language statement includes a variable, an expression to which the variable is set, and a request to determine a data type of the variable. The expression is processed to deduce a data type of the expression. In certain embodiments, the data type of the expression is then assigned as the data type of the variable such that the data type can be used when the programming language statement is compiled into machine executable code.

Abstract: Techniques for predicting failure mode specific reliability characteristics of tangible equipment using parametric probability models are disclosed. In some example embodiments, a computer system receives a model training configuration entered via a user interface, trains a failure curve model for a selected failure mode of a selected equipment model based on the model training configuration at a time indicated by training schedule data, generating, and generates analytical data for the selected failure mode of the selected equipment model using the trained failure curve model. The failure mode corresponds to a specific way in which the equipment model is capable of failing. In some example embodiments, the training of the failure curve model comprises determining a shape parameter and a scale parameter for the failure curve model based on a fitting of failure event data to a continuous probability distribution, and storing the parameters for use in generating the analytical data.

Abstract: Implementations of the present disclosure include receiving, by the centralized sub-system, user data representing a user that is authorized to access one or more electronic documents stored in the centralized sub-system, generating a hash identifier that is unique to the user, the hash identifier being generated based on at least a portion of the user data, and recording, by the decentralized sub-system, the hash identifier in a blockchain that is maintained within the decentralized sub-system, the blockchain being stored in multiple nodes of the decentralized sub-system.

Abstract: Various examples are directed to systems and methods for executing a web application with client-side encryption. A web browser can receive a document comprising a plurality of data elements including a secure element that comprises an encrypted value. An extension component may generate a secure container element to replace the secure element. The extension component can also insert a subdocument into the secure container element. The web browser may be configured to prevent web applications from accessing the subdocument. The extension component may also decrypt the encrypted value to generate a clear value and write the clear value to the subdocument. The web browser may render the document using the clear value.

Abstract: Disclosed is a system and method for improving database memory consumption and performance using compression of time stamp columns. A number of time stamps of a time series is received. The time stamps have a start time, and are separated by an equal increment of time that defines an interval. The start time and interval are stored in a dictionary of a column store of a database. An index is generated in the column store of the database, the index having a number of index vectors. Using the index vectors, each time stamp of the number of time stamps can be calculated from the start time stored in the dictionary and the position in the time series based on the interval stored in the dictionary.

Abstract: Disclosed herein are system, method, and computer program product embodiments for providing paged and compressed storage for column data, while maintaining existing access mechanisms for the data. In order to reduce an in-memory footprint for column data, columns may be stored in pageable format using page chains, and only those pages of the column data needed to resolve a query will be placed in memory, and evicted from memory when no longer needed. In order to further reduce the footprint for these columns, compression can be applied, and the compressed column data stored in the same pageable format using page chains. The compressed data includes a plurality of vectors, each of which is converted into pages and stored on the page chain with the others so that they can be efficiently retrieved during database retrieval operations.

Abstract: Disclosed herein are system, method, and computer program product embodiments for displaying object pages containing data displayed in tables in a client application. The client application may provide a user experience framework via which an administrator may pre-configure the behavior of object pages that display data related to the application in table form. A table may be configured to display using lazy loading, tab navigation, or a preview version to list report based on the nature and expected size of the table. When preview version to list report is chosen, a preview version of a table may be provided to a user to allow the user to interact with the data without displaying the entirety of the dataset.

Abstract: A query is received at a database execution engine. The query is executed by the database execution engine, the executing including executing one or more threads that operate on versions of data of the delta fragment; acquiring and releasing, by a first thread, a latch on a first version of the data of the delta fragment, the releasing is performed prior to the first thread entering a wait state; and reclaiming the first version of the data of the delta fragment from memory and during execution of the query. Related systems, methods, and articles of manufacture are also described.

Abstract: A method for synchronizing customized templates is provided. The method may include detecting a change to a first database table associated with a first cloud-based system hosting a first cloud-based software application. The change may correspond to customizations applied to a first template deployed at the first cloud-based system. The first template may define columns and/or rows in the first database table that is operated upon by the first cloud-based software application. The customizations applied to the first template may be identified based on the first database table. The customizations may be applied to a second template deployed at a second cloud-based system hosting a second cloud-based software application. The second template may define columns and/or rows in a second database table that is operated upon by the second cloud-based software application. Related systems and articles of manufacture, including computer program products, are also provided.

Abstract: In some aspects, there is provided a method including receiving an execution plan file, the execution plan file utilizing at least one operator of interest and further utilizing other actions separate from the at least one operator of interest. The method further includes forming an execution plan object by modifying the execution plan file by isolating the at least one operator of interest from the other actions. The method further includes performing a series of tests executing an extended execution plan object. The series of tests can include receiving the input data identified by the one or more pointers in the extended execution plan object, executing the extended execution plan object using the received input data, measuring, based on the execution of the extended execution plan object, at least one cost metric representative of execution of the at least one operator of interest, and outputting the measured cost metric.

Abstract: The present disclosure relates to computer-implemented methods, software, and systems for generating a hierarchy of metadata tables for a database comprising containers including tables. The tables are identified by table names and assigned to containers. A first table is assigned to two containers and may define two table instances of the first table. The hierarchy of metadata tables includes a first metadata table defining mappings between container identifiers, table names, table sections, and unique identifiers for corresponding data within table sections of table instances defined with the table names mapped to the container identifiers. In response to receiving a request to generate a replication of table content, a second metadata table is generated to identify a unique set of table instances from the set of the containers based on evaluating the first metadata table. The unique set of table instances comprises data from the database storage without repetition.

Abstract: Methods, systems, and computer-readable storage media for determining, by an asset service executing on a server associated with the on-premise system, that a document has been added to the on-premise system, providing, within the cloud-based system, a temporary data object that stores data of the document, generating, within the cloud-based system, a persistent data object based on the temporary data object, the persistent data object storing data of the document, deleting the temporary data object, and associating a document identifier of the document to an asset identifier of the persistent data object within a repository of the on-premise system.

Abstract: Technologies are described for performing automated data integration, reconciliation, and/or self-healing using machine learning. For example, data integration can be checked using a reconciliation procedure. The number of times that the reconciliation is performed can be determined dynamically by a machine learning model. For each iteration, reconciliation can be performed to check integrated data against source data. If any reconciliation errors are found, then self-healing operations can be performed. Results of the reconciliation can be output. The reconciliation results can be used to update the machine learning model so that the machine learning model can dynamically adjust the number of iterations to perform based at least in part on reconciliation results.

Abstract: In an example, information about interactions by users, via the graphical user interface, with a plurality of items rendered in the graphical user interface, is obtained, the information including a timestamp of a time and date at which the corresponding interaction occurred. A series of operations is performed upon detecting a condition that triggers a ranking of items rendered in the graphical user interface, for each item for which the condition applies. A ranking score is calculated for the item and a decay is applied to the ranking score by applying a Gaussian distribution, using the timestamp of the item as an origin to the Gaussian distribution, producing a decayed ranking score. The items for which the condition applies are then ranked based on their respective decayed ranking scores and a subset is rendered in the graphical user interface in an order matching the ranking.