Abstract: The system and method are directed to a prototypical contrastive learning (PCL). The PCL explicitly encodes the hierarchical semantic structure of the dataset into the learned embedding space and prevents the network from exploiting low-level cues for solving the unsupervised learning task. The PCL includes prototypes as the latent variables to help find the maximum-likelihood estimation of the network parameters in an expectation-maximization framework. The PCL iteratively performs an E-step for finding prototypes with clustering and M-step for optimizing the network on a contrastive loss.

Abstract: Disclosed are some implementations of systems, apparatus, methods, and computer program products for facilitating the prediction of the quantity and/or qualities of new data objects of a particular data object type to be generated based upon past generation of data objects of the particular data object type. Data that is used to generate predictions is obtained and filtered according to criteria that are configurable. In some implementations, the criteria indicate an industry for which predictions are generated, a geographic region for which predictions are generated, and/or time period criteria indicating a time period for which the predictions are generated. Predictions may be generated using a computer-generated model, which may be associated with the particular data object type.

Abstract: A system may include an application server and one or more tenants served by the application server. The application server may host a virtual machine with multiple isolated sub-environments. Each tenant of the application server may request to run a program in a tenant-specific sub-environment of the virtual machine. The sub-environments may be configured so the execution of one tenant's code does not affect execution of another tenant's code or the hosting virtual machine, for example, by considering the resources used to execute the code. The application server may implement techniques to securely execute “untrusted” code, programmed using one or more different programming languages, in the sub-environments by enforcing resource restrictions and restricting the sub-environments from accessing the host's local file system. In this way, one tenant's code does not negatively impact execution of another tenant's code by using too many resources of the virtual machine.

Abstract: Techniques are disclosed relating to retrieving data from an in-memory cache, such as that for a database system. In various embodiments, an in-memory cache receives a request from an application for data, where the request specifies a class having a function executable to access the data from a location external to the cache in response to a cache miss. The cache handles the request such that the cache miss is not returned to the application. Specifically, the cache, in some embodiments, determines whether it stores the requested data, and in response to determining that it does not store the data, calls the function of the class to access the data from the location external to the cache and receives the data returned by the execution of the function. The cache then stores the received data in the cache and returns the received data in response to the request.

Abstract: Disclosed herein are system, method, and device embodiments for implementing visual context in dataflows. A computing device may display a dataflow representation associated with a transformative process, receive selection of an individual node of the dataflow representation, and determine a visual context associated with the individual node based on a data transformation step associated with the individual node. Further, the computing device may determine a plurality of nodes associated with the visual context, and apply a graphical effect to the individual node and the plurality of nodes, the graphical effect modifying an attribute of the individual node and the plurality of nodes.

Abstract: A client system may generate a new key pair for a secrets management process. The client may generate a shared secret using the private key of the new key pair and a public key of a secrets management server. Using the shared secret, the client may derive an encryption key and encrypt a data payload for subsequent decryption by the secrets management server. Upon encryption of the data payload, the client may erase the private key. Subsequently, the client or an associated client may call the secrets management server for decryption of the data payload. The secrets management server may derive the encryption key using the public key associated with the encrypted payload and the private key of the secrets management server and use the encryption key to decrypt the data payload for use by the client or an associated client.

Abstract: Database environments may choose to schedule complex analytics processing to be performed by specialized processing environments by caching source datasets or other data needed for the analytics and then outputting results back to customer datasets. It is complex to schedule user database operations, such as running dataflows, recipes, scripts, rules, or the like that may rely on output from the analytics, if the user database operations are on one schedule, while the analytics is on another schedule. User/source datasets may become out of sync and one or both environments may operate on stale data. One way to resolve this problem is to define triggers that, for example, monitor for changes to datasets (or other items of interest) by analytics or other activity and automatically run dataflows, recipes, or the like that are related to the changed datasets (or other items of interest).

Abstract: Embodiments described herein provide systems and methods for generating an adversarial sample with inflectional perturbations for training a natural language processing (NLP) system. A natural language sentence is received at an inflection perturbation module. Tokens are generated from the natural language sentence. For each token that has a part of speech that is a verb, adjective, or an adverb, an inflected form is determined. An adversarial sample of the natural language sentence is generated by detokenizing inflected forms of the tokens. The NLP system is trained using the adversarial sample.

Abstract: Methods for splitting a plurality of records into subsets for parallel processing by separate threads are disclosed. In one embodiment, a plurality of records are added to a database. The adding includes tracking, for each unique value found in the plurality of records within a selected one of the sequence of fields, a statistic based on the values found in the plurality of records within a next one of the sequence of fields. Then a plurality of queries are generated, each includes a row value constructor (RVC) with a logical operation, a specific one of the unique values found within the selected one of the sequence of fields, and, for that specific value, the tracked statistic. Then a plurality of threads are caused to execute that each submit one of the queries and operate on a subset of the plurality of records responsive to execution of that query.

Abstract: In accordance with embodiments, there are provided mechanisms and methods for facilitating criteria-based cost-efficient routing and deployment of metadata packages in an on-demand services environment according to one embodiment. In one embodiment and by way of example, a method comprises analyzing, by a metadata costing server computing device, characteristics of a metadata package awaiting deployment, where analyzing includes matching the characteristics with predefined criteria associated with deployment of metadata packages. The method may further include routing, by the metadata server, the metadata package for real-time deployment if the characteristics satisfy the criteria, where the real-time deployment allows for bypassing of message queues associated with standard deployment.

Abstract: The present disclosure is directed to systems and methods for generating an API caching library using a shared resource file. For example, a method may include: receiving, at a first platform, a shared resource file comprising metadata for declaratively deriving an application programming interface (API) caching library for a native application operating on the first platform and a corresponding application related to the native application for a second platform; parsing the shared resource file to extract the metadata at run-time of the native application; declaratively deriving the API caching library based on the extracted metadata, the declaratively deriving the API caching library comprising creating a plurality of objects that represent respective API endpoints of the API caching library; and executing a function of the native application based on at least one of the API endpoints.

Abstract: A client device may be provisioned with a digital certificate to support various operations. The client may transmit a certificate request to a server. The server may initiate a key agreement process using a short-lived private key generated at the server and a public key of the device to derive a symmetric key. The symmetric key may be used to encrypt a payload that includes the digital certificate and an associated private key. Further, the server initiates a key agreement process using the partial private key that was generated for the client and the short-lived public key. A partial key agreement result, and the encrypted payload may be transmitted to the client. The client may complete the key agreement process using the partial key agreement result and a respective portion of the private key. The client may derive the encryption key and decrypt the payload to access the digital certificate.

Abstract: Estimating a time to perform an operation on a prospective data set of a selected size that includes a plurality of data entities and relationships between the data entities. A number of data sets of different size each comprising a number of like data entities and like relationships between the like data entities are received as input. A number of actions performed on a subset of the number of like data entities and like relationships between the like data entities that substantially comprise the operation are provided as output. For each of the number of data sets of different size, an elapsed time to perform a batch process for each of the number of actions on the subset of the number of like data entities and like relationships between the like data entities that comprise the operation is calculated.

Abstract: Systems, methods, and apparatuses for implementing machine learning models for smart contracts using distributed ledger technologies in a cloud based computing environment are described herein.

Abstract: A method for communicating with a server system, by a client device via a browser application, is provided.

Abstract: Methods, systems, and devices for supporting security for private data inputs to artificial intelligence models are described. A device (e.g., an application server) may receive a request to run an artificial intelligence model. The device may run the artificial intelligence model on a public data set and an extended set of data that includes both the public data set and a private data set. The device may determine a first set of outcomes based on running the artificial intelligence model on the public data set and a second set of outcomes based on rerunning the model on the extended set of data. The device may then compare the two sets of outcomes to determine whether a private data value is identifiable based on the second set of outcomes. If a private data value is identifiable, the device may obfuscate the results prior to transmitting the results to the requestor.

Abstract: Techniques and architectures for creating scratch organizations in a multitenant environment. A scratch organization defined by metadata corresponding to a subject organization is generated. The metadata defines at least a set of privileges for the subject organization to be replicated in the scratch organization. Test data is loaded from a test source that is not the subject organization. Changes to the subject organization are not applied to the scratch organization after creation of the scratch organization and changes to the scratch organization are not applied to the subject organization. One or more test operations are performed on the scratch organization using the loaded test data. The scratch organization is destroyed after the one or more test operations have been performed.

Abstract: A method and system for capturing dynamic data from data fields of a product has been developed. The method includes first collecting a product data set with multiple data fields. Static data fields of the product data set are identified that contain pre-defined standardized data for multiple other data sets. Next, dynamic data fields of the data set are identified. A dynamic data field is created by a user and contains additional data that is unique and applicable only to the product data set. The static data fields and the dynamic data fields are stored separately in a database repository for later recovery.

Abstract: Disclosed embodiments are a computing system and a computer-implemented method related to predicting travel time by using adjacent travel time. The prediction may be based on correcting the nominal predictive travel time estimation using a relation between the predictive travel time estimation and the nominal predictive travel time estimation of a nearby commuting instance in a similar direction. This allows a reduced number of queries to the geographic information system, which may save time, costs and conserve communication resources.

Abstract: Methods and systems for bulk uploading of data in an on-demand service environment are described. In one embodiment, such a method includes retrieving a bulk transaction request from a transaction queue within a host organization, where the bulk transaction request specifies a plurality of database transactions to be processed against a multi-tenant database system of the host organization. Such a method further includes determining a current computational load of the multi-tenant database system, processing a subset of the plurality of database transactions specified by the bulk transaction request against the multi-tenant database system when the current computational load is below a threshold, and updating a bulk transaction results log based on the subset of the plurality of database transactions processed. Remaining database transactions specified by the bulk transaction request may then be re-queued for later processing.