Abstract: In some examples, a system maintains, for a base storage structure and a collection of snapshots associated with the base storage structure, statistics relating to usage of storage space based on data of the base storage structure and the collection of snapshots. The system checks metadata of a first snapshot to determine validity of references in the metadata and a presence of any child of the first snapshot or a grandparent of the first snapshot. The system associates, based on the validity of the references in the metadata and the presence of any child or grandparent of the first snapshot, a hint with a metadata entry of a parent of the first snapshot to indicate that a metadata entry of the child or the grandparent is to be checked. The system updates, using the hint, the statistics relating to usage of storage space.

Abstract: A method for reconstituting a data storage unit of a volume, the method comprising: initializing a reconstituted data storage unit that includes a plurality of portions; identifying a first type-1 data structure that corresponds to the data storage unit and a first snapshot of the volume, the first snapshot being created at a first point-in-time, the first type-1 data structure including a first bitmap, the first bitmap including a first plurality of bits, each of the first plurality of bits corresponding to a different portion of the data storage unit; retrieving, from the first snapshot, one or more portions of the data storage unit that correspond to bits in the first bitmap that are set, and storing the portions that are retrieved from the first snapshot in the reconstituted data storage unit; and returning the reconstituted data storage unit.

Abstract: A method and system for automatic deletion of information based on time synchronization and trusted counting is provided, the method including: defining different data structures including configurations in a normalized manner; setting triggering conditions for automatic information deletion, including conditions for automatic deletion based on time synchronization and/or based on trusted counting; maintaining synchrony of the configurations across different domains; and after satisfying the triggering conditions, making an information source domain and/or an information propagation domain perform the automatic information deletion.

Abstract: A method for performing disaster recovery in a clustered environment comprises identifying, at a master device, a first indexer from a set of indexers to serve as a primary indexer for responding to queries pertaining to a subset of data. The method also comprises assigning, at the master device, a generation identifier indicating that the first indexer is the primary indexer for the subset of data. Responsive to an event prompting a change in a primary indexer designation for the subset of data, the method comprises identifying, at the master device, a second indexer from the set of indexers to serve as the primary indexer for responding to queries pertaining to the subset of data. Further, the method comprises assigning, at the master device, a new generation identifier indicating that the second indexer is the primary indexer for the subset of data.

Abstract: A data processing method and system for a distributed database, and a device and a storage medium. The method comprises: sorting transactions in a transQueue by using a global timestamp TSO, creating a local transaction list by means of an XA Event and a waitTrans in the transQueue, and generating first-level sorting queues according to the local transaction list (102); for the first-level sorting queue of each DN, performing multi-path merging and sorting, so as to generate a global sorting queue (104); and for the global sorting queue, merging local transactions that have the same global timestamp TSO, and generating a global binary log Binlog (106). Therefore, the global orderliness and integrity of transactions during the process of processing and replicating data in a distributed database are ensured.

Abstract: Methods and systems for managing access and control of data are disclosed. To manage access and control, a data management system may use a variety of data minimization and resiliency techniques to differentiate different repositories for storage of data. The different repositories may store data in different manners, and provide different levels of performance regarding stored data. By differentiating the repositories, criteria may be established and enforced to prioritize different types of data for storage.

Abstract: Disclosed is a method and mapping server for generating mapping information for a cloud event. The method includes: obtaining a document for any one mapping target event; generating at least one of a topic and a keyword for the mapping target event based on the document; based on at least one of the topic and the keyword, selecting one tactic to which the mapping target event is mapped; based on at least one of the topic and the keyword, selecting a technique to which the mapping target event is mapped; storing mapping information for the mapping target event in a mapping dictionary; and generating mapping information for each event and storing the generated mapping information in the mapping dictionary, thereby generating a mapping dictionary for at least some of the plurality of events.

Abstract: A method for performing hierarchical classification of data is disclosed. The method is being executed by at least one processing device. The method includes receiving input data for encoding into multiple channels. Further, the method includes extracting one or more features and one or more temporal structures corresponding to the input data. The method further includes identifying one or more feature dependences in the input data. Further, the method includes combining the extracted one or more features corresponding to the input data, the extracted one or more temporal structures of the input data, and the identified one or more feature dependencies in the input data into a combined feature set. Thereafter, the method includes classifying the combined feature set into one or more output classes, and thereby performing the hierarchical classification of data.

Abstract: The technology described herein is directed to generating prompts for a language model, such as a large language model. In this regard, an enterprise interpretation model may receive a request for assistance with an enterprise application record. In response to receiving the request for assistance, the enterprise interpretation model may determine data corresponding to the record. Based on the determined data, the enterprise interpretation model may generate a query corresponding to the record. The query may include at least a subset of the data corresponding to the record. The query may then be provided to the language model for a response to be generated.

Abstract: Systems and methods presented herein provide means to process a target text, such as a medical record, in search of keyword(s). A set of medical records are accessed, where the set of medical records includes the target text. A set of target keywords are initialized. One or more natural language processing (NLP) models are used to expand the set of target keywords based on the initialized set of keywords. The one or more NLP models process the target text with the expanded set of target keywords to extract a set of found keywords. A report including the set of found keywords and at least one text excerpt from the target text is caused for display.

Abstract: Provided in the present disclosure are a text recommendation method and apparatus, a model training method and apparatus, and a readable storage medium. The text recommendation method includes: acquiring text retrieval information from a user; when it is determined that there is historical text retrieval information for the user, determining text information of each text in a text set retrieved by using the text retrieval information; performing embedded representation on the text information of each text based on a self-attention model, and determining a text embedding vector of each text; inputting the text embedding vector of each text into a trained graph convolutional network model, to obtain the probability of interaction between the user and each text in the text set; and screening out, from the text set, target text which meets a preset interaction probability, and recommending the target text to the user.

Abstract: Systems and methods for collecting and processing memory-related data and synthesizing memory-based narratives that includes collecting memory input data from one or more users, constructing a memory graph comprising a hierarchical data structure including one or more graphical nodes for storing the memory input data, assimilating the memory graph into a semantic collage of distinct memory graphs, generating a mnemonic narrative artifact comprising memory input data from related memory graphs, and surfacing the mnemonic narrative artifact via a user interface to one or more users.

Abstract: System, method and interface for interpreting natural language comparisons during visual analysis are provided. The system includes obtaining a natural language utterance that includes a comparison query and a dataset of attributes and values relevant to interpreting the comparison query. The system also includes interpreting the natural language utterance based on the dataset using multi-step chain-of-thought reasoning prompting to generate a response to the comparison query. The system also includes generating a visualization based on the response and a text summary describing the multi-step chain-of-thought reasoning for the comparison query.

Abstract: The present disclosure is directed toward systems, methods, and non-transitory computer readable media for generating and managing multilocational data blocks, generating and summarizing content blocks within a virtual space interface, and generating and providing a content block browser as part of a virtual space platform. In some embodiments, the disclosed systems generate a multilocational data block that includes a block identifier that is tied to a source identifier for embedding digital content from a network location indicated by the source identifier. The disclosed systems can also generate block summaries from content blocks for presenting and modifying digital content embedded within the content blocks via block identifiers and source identifiers. In some embodiments, the content block system can provide a content-block-based web browser in the form of a virtual space that includes embedded content blocks that integrate webpage functionality.

Abstract: In a state in which the language data in a tree structure includes at least one node and the at least one node includes at least one word, a plurality of word vectors including a first word vector and a second word vector is generated by a management server, based on the number of words included in each of a plurality of pieces of language data, and a dot product function of the first word vector and the second word vector is used by the management server to measure a score of similarity between first language data corresponding to the first word vector and second language data corresponding to the second word vector so that language data is managed for determining similarity by a management server.

Abstract: In some implementations, a labeling device may determine a label value associated with a label of the document that is associated with a first application. The labeling device may generate a label map that includes an indication of a key-value pair associated with the label value. The labeling device may write, based on the key-value pair associated with the label value, the label value to one or more file properties associated with the document, or may add, based on the key-value pair associated with the label value, the label value as content associated with the document. The label value may persist with the document based on transferring the document from the first application to a second application.

Abstract: Methods, systems, and apparatus are provided for generating an image. A personalized text prompt is generated by processing an input embedding using a transformer model followed by a first fully connected neural network. The input embedding comprises a multi-dimensional embedding vector associated with a user profile and a plurality of user items. A scored label set is generated identifying a user's preferences by processing a set of attributes for the plurality of user items using a second fully connected neural network. The image is generated by processing the personalized text prompt and the scored label set using a diffusion model.

Abstract: A method including pre-screening one or more second images from a database for a search result based on one or more substring distances between one or more first binary substrings for a first image and one or more second binary substrings for the one or more second images, comprises: determining the one or more substring distances between one or more substring pairs of the one or more first binary substrings and the one or more second binary substrings of the one or more second images, and upon determining that the one or more substring distances and the one or more second binary substrings are not greater than one or more substring distance thresholds, including the one or more second images in the search result. The method further can include after pre-screening, determining one or more image distances for one or more third images of the search result.

Abstract: A text-based search optimization via implicit image search augmentation eliminates or reduces the need for providing an image query input, performing multiple search queries, displaying multiple user interfaces, and the like by enabling a search engine to return a single set of search results comprising an aggregated and ranked set of text-based results and a set of image-based results based on one or more text-based keywords of a search query. Initially, a search query comprising one or more text-based keywords is received at a search engine. A machine learning model is utilized to generate an image based on a first portion of the one or more text-based keywords. Image-based results are generated based on the image. Text-based results are generated based on a second portion of the one or more text-based keywords. The image-based results and the text-based results are aggregated and ranked in a single set of search results.

Abstract: An apparatus for processing fingerprint hashes of an audio stream. The apparatus includes: hardware caches, each hardware cache configured to store files, each file containing a part of a fixed size continuous data structure, and the fixed size continuous data structure containing fixed size data items, each fixed size data item representing a stored unique fingerprint hash; random access memories; and processors. The processors are configured to cause performance of: obtaining generated fingerprint hashes of an audio stream; matching recursively each generated fingerprint hash of the audio stream against stored unique fingerprint hashes of the tracks using match time offsets between the audio stream and each track in order to identify the audio stream; wherein the matching comprises retrieving, into the random access memories from one of the files, each stored unique fingerprint hash addressed by each generated fingerprint hash and represented by each fixed size data item.

Abstract: Systems and methods for cross-modal retrieval are provided. According to one aspect, a method for cross-modal retrieval includes obtaining a query describing a sound using a query modality other than a sound modality; encoding the query to obtain a query embedding using a query encoder network for the query modality and a query projection network, wherein the query projection network includes a self-attention layer, and wherein the query embedding is in a joint embedding space for the query modality and the sound modality; and providing a response including an audio sample based on the query embedding, wherein the audio sample includes the sound.

Abstract: Provided are a method and an apparatus for lyric video display, an electronic device, and a computer-readable medium. The method includes: acquiring, based on a lyric video display operation of a user, multimedia data to be displayed, the multimedia data including audio data and lyrics; performing a segmentation processing on the lyrics to obtain segmented lyric fragments; determining a target time point; and acquiring, from the segmented lyric fragments, a target lyric fragment corresponding to the target time point; and displaying the target lyric fragment in combination with a background content in a background area, and playing audio data corresponding to the target lyric fragment.

Abstract: Systems, apparatus, articles of manufacture, and methods are disclosed. An example system includes programmable circuitry; a memory that stores executable instructions that, when executed or instantiated by the programmable circuitry, facilitate performance of operations, comprising: forming a data structure representing a request to a content provider for a media stream; modifying the data structure to represent a request to a local manager for the media stream, the modified request data structure including an address corresponding to the local manager; transmitting a request to the local manager as a hypertext transfer protocol (HTTP) request, the request based on the modified data structure; and receiving a response from the local manager as a user datagram protocol (UDP) stream.

Abstract: A video analysis system receives one or more queries from users of client devices. The video analysis system trains a machine-learned video encoder and/or a decoder coupled to receive video data and a prompt including a user query and generate an output for responding to the user query. A set of video embeddings are generated by extracting frame data, audio data, or text data from the video content, and applying a machine-learned video encoder to the frame data, the audio data, or the text data to generate the set of video embeddings. The video analysis system also generates a set of prompt embeddings representing at least a portion of the query in a latent space. The video analysis system applies at least a component of a machine-learned decoder to the input tensor to generate an output including a set of output embeddings.

Abstract: Example techniques for using configurable functions to harmonize data from disparate sources may include retrieving first and second datasets including data records with values for respective first and second sets of fields; analyzing the first and second sets of fields to identify a third set of fields included in both the first and second sets of fields; identifying data records from the first and second plurality of data records having matching values for the third set of fields; stitching the identified data records from the first and second pluralities of data records with one another in order to generate a third dataset including a third plurality of data records having values for both the first and second set of fields; applying one or more functions to the third plurality of data records of the third dataset to produce an output dataset; and displaying the output dataset via a user interface.

Abstract: Systems and methods are described for procedurally-based decision support using ontology-based classification of database records. The procedural decision support can facilitate extraction of contextually relevant data from a database. The data may be formatted for compatibility with a knowledge-based data library using one or more scripts and populated in the library as an entity. Classification of the entity can be reasoned using the available data. One or more classifications of the entity may be returned to the procedural decision support to facilitate computation of a recommendation.

Abstract: According to embodiments of the disclosure, a method and an apparatus for information processing are provided. The method includes determining a plurality of pending events from one or more media contents, at least two of the plurality of pending events having an event relationship. The method further includes determining an additional event set from the plurality of pending events based on an existing event set for constructing an event evolutionary graph, an additional event in the additional event set being different from an existing event in the existing event set. The method further includes updating the event evolutionary graph based on the additional event set. In this way, incremental updates of the event evolutionary graph can be realized, thereby improving the construction efficiency of the event evolutionary graph.

Abstract: A document management system can include an artificial intelligence-based document manager that can perform one or more predictive operations based on characteristics of a user, a document, a user account, or historical document activity. For instance, the document management system can apply a machine-learning model to determine how long an expiring agreement document is likely to take to renegotiate and can prompt a user to begin the renegotiation process in advance. The document management system can detect a change to language in a particular clause type and can prompt a user to update other documents that include the clause type to include the change. The document management system can determine a type of a document being worked on and can identify one or more actions that a corresponding user may want to take using a machine-learning model trained on similar documents and similar users.

Abstract: An automated software-as-a-service (SaaS) security posture management (SSPM) system disclosed herein detects and maintains security posture for SaaS applications according to correct implementation of configuration settings. Based on detecting a previously unseen SaaS application with unknown implementation of configuration settings, the SSPM system scrapes the Internet for web content for the SaaS application and preprocesses/inputs the web content into a machine learning model to obtain predictions of correct/incorrect implementation of configuration settings as output. Based on the predictions not having sufficiently high confidence, the SSPM system obtains additional application content by logging into the SaaS application and scraping locally rendered pages therein. The application content is preprocessed/input to the machine learning model to obtain additional high confidence predictions.

Abstract: A method and system are disclosed for providing treasury management information to a user through a web-based natural language interface in communication with a secure financial and treasury management platform. The primary system components are a priming engine that prompts a large language model (LLM) with a user's profile and history data, an action engine that prompts the LLM based on natural language queries from the user, and an indexed interactive financial platform that calculates financial data requested by the user. The action engine may also interact with application programming interfaces in support of performing financial and treasury management operations requested by the user.

Abstract: A multimodal search system is described. The system can receive image data captured by a camera of a user device. Additionally, the system can receive audio data associated with the image data. The audio data can be captured by a microphone of the user device. Moreover, the system can process the image data to generate visual features. Furthermore, the system can process the audio data to generate a plurality of words. The system can generate a plurality of search terms based on the plurality of words and the visual features. Subsequently, the system can determine one or more search results associated with the plurality of search terms and provide the one or more search results as an output.

Abstract: Disclosed herein are system, method, and computer program product embodiments for generating intelligent prompts. An embodiment operates by identifying a plurality of content elements in a content object retrieved from a third-party system. The embodiment receives supplemental data collected by one or more enricher micro services invoked based on the plurality of content elements in the content object. The embodiment identifies one or more topics in the content object. The embodiment further creates a nudge card for the content object that includes the plurality of content elements, the supplemental data, the one or more topics, an expiration timer, and at least one interface element associate with an action. The embodiment transmits the nudge card for the content object within a nudge card feed to a client device for presentation to a user.

Abstract: A method for communication between a terminal and a communication equipment, via a communication network, is described. The method includes transmitting, from the terminal to the equipment, via the network, a search request for a public place, the request including the type of place and at least one word characterizing the type, the at least one word being associated with a first intensity value assigned to at least one human sense. The method also includes selecting, in the network, one or more public places for which a second intensity value for said at least one sense has been assigned and the second value of which is closest to the first value, and, in response to the request, receiving at the terminal a content item from the equipment including identification information in relation to the selected place/places.

Abstract: Methods and systems for identifying a meaning of an ambiguous term in a natural language query. The media guidance application isolates first and second terms from a query received from a user and identifies, in a knowledge graph, first and second pluralities of candidate components associated with the first and second terms. The first and second terms each having multiple candidate components indicates the first and second terms have ambiguous meanings. The media guidance application matches each candidate component of the first and second pluralities of candidate components to form a plurality of pairs and determines strength of association for each pair in the plurality of pairs. The media guidance application filters the plurality of pairs by strength of association for each pair and determines a plurality of possible meanings based on the filtered plurality of pairs. The media guidance application selects a meaning from the plurality of possible meanings.

Abstract: A method for changing an itinerary based on a user change request is disclosed. The method may commence with receiving an itinerary request associated with one or more passengers. The method may continue with receiving a user itinerary change request associated with the itinerary network. The method may continue with generating an itinerary object associated with the user itinerary change request. The method may continue with modifying the itinerary network based on the itinerary object. The method may continue with processing the itinerary network using a topology of the itinerary network to create a plurality of tuples, the plurality of tuples including at least flight tuples and hotel tuples. The method may continue with performing a content search for the plurality of tuples for each of the one or more passengers. The method may continue with generating feasible itinerary solutions based on results of the content searches.

Abstract: The discussion relates to providing more relevant search results to a user based upon information relating to IoT devices associated with the user. One example can receive search results for a search query associated with the user. The example can obtain entities from IoT devices associated with the user and rank the search results utilizing the entities from the IoT devices.

Abstract: A computing system obtain a keyword and an identifier for a user of a content authoring application. Based upon the keyword and identifier for the user, the computing system walks a user graph comprising nodes connected by edges. The walk comprises identifying seed nodes in the user graph representing at least one topic that corresponds to the keyword and identifying second level nodes in the user graph that are connected to the seed nodes. The second level nodes represent first content that is associated with the user. The computing system transmits contextual data that is based upon the first content to the content authoring application. The contextual data is processed and formatted and is included in second content presentable by the content authoring application. The contextual data may be used to autogenerate the second content without user input. The second content may be modified by the user as desired.

Abstract: Techniques for an artificial intelligence (AI) platform to search a document collection are described. Embodiments may use AI and machine learning techniques within a framework of an electronic document management system to perform semantic searching of an electronic document or a collection of electronic documents for certain types of information. The AI platform may summarize the information in a natural language representation of a human language. Other embodiments are described and claimed.

Abstract: Methods, systems, and computer-readable media are provided for collective reconciliation. In some implementations, a query is received, wherein the query is associated at least in part with a type of entity. Top-ranked search results are caused to be generated based at least in part on the query. An entity of the type of entity is identified in the top-ranked results and a response to the query is generated from the top-ranked results, the response including a reference to the entity that includes text distinct from the terms of the query.

Abstract: An action selection facility for a website made up of pages is described. The facility receives requests, each identifying a requested page among the website's pages, and a requesting user for whom the request was generated. For each received request, the facility: (a) identifies actions presently available to the requesting user that can be performed using the website; (b) accesses a set of action determinant factors that relate to the request, and uses them to generate a score for each identified action; and (c) selects a proper subset of the identified actions based at least in part on the scores generated for them.

Abstract: In some implementations, a tag library device may receive, from first user devices, an indication of a set of tags with a corresponding set of properties and a corresponding set of rules. The tag library device may build a data structure that stores the set of tags in association with the corresponding set of properties and the corresponding set of rules and may transmit the data structure to second user devices such that a portion of the set of tags are included in a web site published by the second user devices. The tag library device may receive, from a platform associated with the web site, an indication of a set of events associated with the web site. The tag library device may determine statistics based on the set of events and may transmit a report including the statistics to third user devices associated with analytics for the web site.

Abstract: Provided herein are system, method and/or computer program product embodiments, and/or combinations and sub-combinations thereof for controlling access to a website. In some embodiments, a user scans a machine-readable indicia with their mobile device to access a backend redirect page. The backend redirect page receives the access request, generates a unique session identifier, and configures a dynamic address based on the unique session identifier. The backend redirect page redirects the mobile device to an intake page of a website using the dynamic address for that page. Whenever the user attempts to load a page in the website using the corresponding dynamic address, an intake backend for frontend verifies the validity of the unique session identifier associated with the dynamic address. A session monitor tracks the amount of time since the unique session identifier was generated and invalidates the unique session identifier after a predetermined amount of time has passed.

Abstract: A method and apparatus of processing a user initiated request for information is disclosed. The method may provide receiving a user initiated request including a uniform resource locator (URL) submitted to a web browser application and receiving a response that the URL is an invalid URL that cannot be resolved to a corresponding webpage. The method may also include blocking a subsequent attempt to transmit the user initiated request as a browser modified search request that includes modifications to the user initiated request.

Abstract: Implementations relate to determining whether and/or how to implement a user request to prevent a particular search result from being provided in response to a search query. Some of those implementations grant or deny the request based on processing of the particular search result, the search query, and/or account information for a user submitting the user request. For example, some implementations process such information utilizing a classifier in determining whether to automatically deny the request, automatically approve the request, or to provide the request for manual review. Some additional or alternative implementations at least selectively automatically expand (or suggest for automatic expansion) an approval of a request to search result(s) and/or to one or more search queries that are not specified in the request.

Abstract: This application discloses a page rendering method performed by an electronic device. The method includes: parsing a target page to obtain a corresponding page configuration tree; constructing a layout node tree of the target page using the page configuration tree; obtaining layout information of the target page by traversing nodes in the layout node tree of the target page using a cross-platform layout system; generating a painting instruction based on the layout information of the target page and a render object tree; and painting the target page in response to the painting instruction.

Abstract: The invention provides a browser side virtual path rewriting method, system and computer device, including: the system generates first level sites of a portal based on user role permissions, and loads all site pages and sub sites within the permissions according to the first site ID of the first level sites; when the browser renders the generated page, run the script file instructions to replace the original access address of the browser with the main path to be rewritten. After the user clicks a site, the script instruction calculates the path from the site to the first level site, and forms a sub site path by all site names and the separators between sites on the path; Call the history object, run the script command, and replace the original access address of the browser with the main path/site path to be rewritten.

Abstract: One example method includes accessing a parameter space including a set of binary inputs for an unconstrained objective function. The set of binary inputs are solved at a CPU or GPU using an algorithm that is different from the unconstrained objective function to generate a target solution. A subset of the binary inputs is selected. The unconstrained objective function is solved using the selected subset of binary inputs to generate a solution for each of the selected subset of binary inputs. A maximum possible change for each of the selected subset of binary inputs is determined. The maximum possible change defines a subspace including related binary inputs that are located around each of the selected subset of binary inputs. Those binary inputs and their corresponding related binary inputs whose solutions are greater than the target solution are removed from the parameter space to thereby generate a reduced parameter space.

Abstract: Techniques for reduced latency tensor transposition without using a redundant buffer are enabled. Reads from and writes to a buffer array may occur in different dimensions in a neural processing unit (NPU). For example, a set of tensor vectors may be written to a buffer in columnar format and read from the buffer in row format. As vectors of a first tensor are read from the buffer, incoming vectors from a second tensor may be transposed for storage in the dimension of already-read vectors without overwriting unread vectors. Write and read operations may alternately transpose vectors for continuous buffering in a single buffer with reduced latency.

Abstract: Disclosed herein are an outer product-based matrix-vector multiplication operation apparatus and a method using the same. The outer product-based matrix-vector multiplication operation apparatus includes internal calculators, each configured to generate an accumulated value by performing a Multiply-Accumulation (MAC) operation, an internal data transmission path configured to simultaneously provide a vector element to two or more internal calculators, and at least one multiplexer configured to select any one of the vector element and a vector of a matrix, wherein a first input port of each of the internal calculators is connected to one of vectors of the matrix and a second input port of each of the internal calculators is connected to the vector element.

Abstract: Methods, systems, and apparatus for a matrix multiply unit implemented as a systolic array of cells are disclosed. The matrix multiply unit may include cells arranged in columns of the systolic array. Two chains of weight shift registers per column of the systolic array are in the matrix multiply unit. Each weight shift register is connected to only one chain and each cell is connected to only one weight shift register. A weight matrix register per cell is configured to store a weight input received from a weight shift register. A multiply unit is coupled to the weight matrix register and configured to multiply the weight input of the weight matrix register with a vector data input in order to obtain a multiplication result.