Abstract: A method includes defining a set of context types; defining a set of source types, each comprising context types; defining, for each source type, and for each context type included in the events from data sources having the source type, a context definition comprising a set of fields, in events from the data sources, that are associated with the context type; receiving a query comprising a first field value and a time period; retrieving a plurality of events that include the first field value and the time period; for each retrieved event, and for each context definition defined for a source type and a context type of a data source from which the retrieved event originated, determining field values of fields in the set of fields of the context definition; aggregating, for each context type, determined field values from the events; and generating an output.

Abstract: A computerized-method for scaling automatic deployment of a machine-learning detection model in a cloud-based managed analytics service by knowledge sharing to overcome an imbalanced dataset learning problem. The computerized-method includes: sending the received data to machine-learning models to synthesize patterns of the received data to yield a differential privacy data; maintaining in the database the differential privacy data of one or more on-prem cloud-based managed analytics services to generate a consortium shared synthetic data lake; operating phases of machine-learning detection model based on the received data and data in the database to create a packaged model. The data in the database is aggregated and used during the operating phases of the machine-learning detection model to create a packaged model for other on-prem cloud-based managed analytics services, thus overcoming imbalanced dataset learning thereof, and after the packaged model is created it is automatically deployed on-prem.

Abstract: Methods and systems for identifying areas of interest in an image and management of images are disclosed. To manage identification of areas of interest in an image, subject matter expert driven processes may be used to identify the areas of interest. The identified areas of interest may be used to maintain a database usable to guide subsequent use of the images. The database may associate image segments of the images with various landmarks and/or area of interest. The associations may be used to limit the quantity of image segments read from storage during subsequent use of the images.

Abstract: An inference system applies a machine-learning transformer model to a batch of requests with variable input length or variable target length or variable internal state length by selectively batching a subset of operations in the transformer model but processing requests in the batch individually for a subset of operations in the transformer model. In one embodiment, the operation to be processed individually is an attention operation of an encoder or a decoder of the transformer model. By selective batching, the inference system can allow batching operations to be performed for a batch of requests with variable input or target length or internal state length to utilize the parallel computation capabilities of hardware accelerators while preventing unnecessary computations that occur for workarounds that restrain the data of a batch of requests to a same length.

Abstract: To reduce a calculation processing load as a whole while realizing a small amount of data loss for at least one of compression and decompression. For each of a plurality of pieces of data, a storage system determines a compression operation scale of the data based on a feature of the data, executes a lossy compression operation according to the determined compression operation scale to covert the data into encoded data, and stores the encoded data or compressed data thereof into a storage device.

Abstract: The present invention refers to the field of transactional data management. The invention provides a transactional semantic concurrency control system for managing a database, wherein the system comprises a data store engine which is adapted to manage the database comprising value rows, and each value row comprising a primary key uniquely identifying each row, at least one semantic column and a commit timestamp; wherein the data store engine is further adapted to manage operation rows, which are rows comprising a primary key, at least one operation to be applied over at least one semantic column of a value row identified by the primary key, and a commit timestamp; store the value rows and the operation rows; generate an operation row when it receives an operation that inserts, updates, upserts, and/or deletes a row of the database with the commit timestamp of the operation; and apply the operations of at least one operation row when a trigger condition is satisfied.

Abstract: An inference system applies a machine-learning transformer model to a batch of requests with variable input length or variable target length or variable internal sate length by selectively batching a subset of operations in the transformer model but processing requests in the batch individually for a subset of operations in the transformer model. In one embodiment, the operation to be processed individually is an attention operation of an encoder or a decoder of the transformer model. By selective batching, the inference system can allow batching operations to be performed for a batch of requests with variable input or target length or internal state length to utilize the parallel computation capabilities of hardware accelerators while preventing unnecessary computations that occur for workarounds that restrain the data of a batch of requests to a same length.

Abstract: According to principles described herein, a system applies Active Learning methodology to multiple models simultaneously. The system includes a means to distribute the sample selection algorithm across large pools of unlabeled data and a automatic model training deployed on hardware matched to the model type that scales to large volumes of data without consuming all resources.

Abstract: Provided is an edge compute platform (“ECP”) for serving optimized content from local cache or from output of a shared customizable function executed by a compute device at the network edge on behalf of different customer content such that the function is not redundantly deployed for different customer content, and is not be executed each time the same variant of the optimized content is requested. The ECP may canonicalize first transformation parameters of a received original request according to a transformation parameter definition of a particular function that is implicated by the original request, may generate second transformation parameters with a different ordering than the first transformation parameters as a result of the canonicalization, may generate a variant of the original file by inputting the second transformation parameters to the particular function, and may provide the variant in response to the original request.

Abstract: A method of processing a query to a database from a query source is provided, comprising: receiving the query, the query in a first format supported by the query source; inputting the query into a first neural network; outputting, by the first neural network, the query in a second format, wherein the second format is a format supported by the database; receiving, from the database, a response to the query, the response in the second format; inputting the response to the query into a second neural network; outputting, by the second neural network, the response to the query in the first format; wherein each neural network is trained by inputting a first plurality of pairs of semi-structured data, each pair of semi-structured data comprising a sample query or response in the first format and the sample query or response in the second format.

Abstract: A cloud resource join query for join operations across cloud resources is parsed to extract join rules and queries to each cloud resource in the cloud resource join query. Results from the individual cloud queries are dynamically indexed based on pairs of cloud resources indicated in the join rules. A search engine applies first order predicates in the join rules using the dynamic indexes to generate pairwise join results corresponding to the query. A result for the cloud resource join query comprises the pairwise join results after merging.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for searching an autonomous vehicle sensor data repository. One of the methods includes maintaining a collection of sensor samples and, for each sensor sample, an embedding of the sensor sample; receiving a request specifying a query sensor sample, wherein the query sensor sample characterizes a query environment region; and identifying, from the collection of sensor samples, a plurality of relevant sensor samples that characterize similar environment regions to the query environment region, comprising: processing the query sensor sample through the embedding neural network to generate a query embedding; and identifying, from sensor samples in a subset of the sensor samples in the collection, a plurality of sensor samples that have embeddings that are closest to the query embedding.

Abstract: Embodiments of the present invention provide systems, methods, and computer storage media for latent summarization of a graph. Structural features can be captured from feature vectors associated with each node of the graph by applying base functions on the feature vectors and iteratively applying relational operators to successive feature matrices to derive deeper inductive relational functions that capture higher-order structural information in different subgraphs of increasing size (node separations). Heterogeneity can be summarized by performing capturing features in appropriate subgraphs (e.g., node-centric neighborhoods associated with each node type, edge direction, and/or edge type). Binning and/or dimensionality reduction can be applied to the resulting feature matrices.

Abstract: Techniques are described for enabling analysts and other users of an IT operations platform to identify certain data objects managed by the platform (for example, events, files, notes, actions results, etc.) as “evidence” when such data objects are believed to be of particular significance to an investigation or other matter. For example, an event generated based on data ingested from an anti-virus service and representing a security-related incident might include artifacts indicating an asset identifier, a hash value of a suspected malicious file, a file path on the infected endpoint, and so forth. An analyst can use various interfaces and interface elements of an IT operations platform to indicate which of such events and/or artifacts, if any, represent evidence in the context of the investigation that the analyst is conducting. In response, the IT operations platform can perform various automated actions.

Abstract: Novel tools and techniques are provided for implementing an imaging discovery utility for augmenting clinical image management. In some embodiments, in response to receiving a request for a first medical image file(s) from a requesting device, a non-relational (“NoSQL”) data management system (“DMS”) may access a NoSQL database containing, inter alia, a plurality of medical image files that are mirrored copies of a plurality of medical image files stored in a relational (“SQL”) database, the medical image files each being organized in an image-centric hierarchy with image data being at a top level and patient information associated with the image data being at a lower level. Based on a successful search of the NoSQL database based on search terms in the request, the NoSQL DMS may identify a corresponding second medical image(s) in the SQL database, and may retrieve and send (to the requesting device) the identified second medical image(s).

Abstract: Certain embodiments involve a method for generating a search result. The method includes processing devices performing operations including receiving a query having a text input by a joint embedding model trained to generate an image result. Training the joint embedding model includes accessing a set of images and textual information. Training further includes encoding the images into image feature vectors based on spatial features. Further, training includes encoding the textual information into textual feature vectors based on semantic information. Training further includes generating a set of image-text pairs based on matches between image feature vectors and textual feature vectors. Further, training includes generating a visual grounding dataset based on spatial information. Training further includes generating a set of visual-semantic joint embeddings by grounding the image-text pairs with the visual grounding dataset.

Abstract: Computerized systems are provided for detecting one or more tables and performing information extraction and analysis on any given table. Information can be extracted from one or more cells or fields of a table and feature vectors representing individual cells, rows, and/or columns of the table can be derived and concatenated together. In this way, embodiments can use some or all of the “context” or values contained in various feature vectors representing some or all of a single table as signals or factors to consider when generating a decision statistic, such as a classification prediction, for a particular cell.

Abstract: An example system includes a processor to receive a query. The processor can retrieve ranked candidates from an index based on the query. The processor can re-rank the ranked candidates using a Bidirectional Encoder Representations from Transformers (BERT) query-question (Q-q) model trained to match queries to questions of a frequently asked question (FAQ) dataset, wherein the BERT Q-q model is fine-tuned using paraphrases generated for the questions in the FAQ dataset. The processor can return the re-ranked candidates in response to the query.

Abstract: An initial software scan is performed to detect a set of software deployed on an endpoint. An ongoing scan of the endpoint is performed to map a set of file directories associated with each software within the set of software. Via the ongoing scan, a usage frequency for each mapped file directory is determined. A heat map is generated for each mapped file directory, according to usage frequency, using a randomized meta-heuristic. A request is received for a software discovery scan result. The software discovery scan, based on the heat map, is performed in response to the request. The result of the software discovery scan is provided to a user.

Abstract: In some examples, dark web content analysis and identification may include ascertaining data that includes text and images, and analyzing the data by performing deep learning based text and image processing to extract text embedded in the images, and deep embedded clustering to generate clusters. Clusters that are to be monitored may be ascertained from the generated clusters. A determination may be made as to whether the ascertained data is sufficient for classification. If so, a deep convolutional generative adversarial networks (DCGAN) based detector may be utilized to analyze further data with respect to the ascertained clusters, and alternatively, a convolutional neural network (CNN) based detector may be utilized to analyze the further data with respect to the ascertained clusters. Based on the analysis of the further data, an operation associated with a website related to the further data may be controlled.